Ingredients for protist coexistence: competition, endosymbiosis and a pinch of biochemical interactions


  • Johann P. Müller,

    Corresponding author
    1. UFR Sciences de la Vie, UPMC Univ Paris 06, Paris, France
    2. Laboratoire Bioemco, Ecole Normale Supérieure, Paris, France
      Correspondence author. E-mail:
    Search for more papers by this author
  • Céline Hauzy,

    1. IFM Theory and Modelling, Linköping University, Linköping, Sweden
    2. Université Pierre et Marie Curie, Sorbonne Universités, UMR 7625 Ecologie et Evolution, Paris, France
    3. INRA, USC 2031 Ecologie des Populations et Communautés, Paris, France
    Search for more papers by this author
  • Florence D. Hulot

    1. UFR Sciences de la Vie, UPMC Univ Paris 06, Paris, France
    2. Laboratoire d’Ecologie, Systématique & Evolution, UMR 8079, Univ Paris-Sud, Orsay, France
    Search for more papers by this author

Correspondence author. E-mail:


1. The interaction between mutualism, facilitation or interference and exploitation competition is of major interest as it may govern species coexistence. However, the interplay of these mechanisms has received little attention. This issue dates back to Gause, who experimentally explored competition using protists as a model [Gause, G.F. (1935) Vérifications expérimentales de la théorie mathématique de la lutte pour la vie. Actualités Scientifiques et Industrielles, 277]. He showed the coexistence of Paramecium caudatum with a potentially allelopathic species, Paramecium bursaria.

2.Paramecium bursaria hosts the green algae Chlorella vulgaris. Therefore, P. bursaria may benefit from carbohydrates synthesised by the algae. Studying endosymbiosis with P. bursaria is possible as it can be freed of its endosymbiont. In addition, C. vulgaris is known to produce allelochemicals, and P. bursaria may benefit also from allelopathic compounds.

3. We designed an experiment to separate the effects of resource exploitation, endosymbiosis and allelopathy and to assess their relative importance for the coexistence of P. bursaria with a competitor that exploits the same resource, bacteria. The experiment was repeated with two competitors, Colpidium striatum or Tetrahymena pyriformis.

4. Results show that the presence of the endosymbiont enables the coexistence of competitors, while its loss leads to competitive exclusion. These results are in agreement with predictions based on resource equilibrium density of monocultures (R*) supporting the idea that P. bursaria’s endosymbiont is a resource provider for its host. When P. bursaria and T. pyriformis coexist, the density of the latter shows large variation that match the effects of culture medium of P. bursaria. Our experiment suggests these effects are because of biochemicals produced in P. bursaria culture.

5. Our results expose the hidden diversity of mechanisms that underlie competitive interactions. They thus support Gauses’s speculation (1935) that allelopathic effects might have been involved in his competition experiments. We discuss how a species engaged both in competition for a resource and in costly interference such as allelopathy may counterbalance these costs with a resource-provider endosymbiont.


Understanding species coexistence is a seminal issue in ecology. It is well known that the interplay of interspecific interactions such as competition and predation may allow species to coexist (Chase et al. 2002; Holt 1984; Paine 1966). But there is also a growing theoretical and experimental body of evidence for the structural roles played by mutualism, facilitation or interference in interaction with exploitation competition (Amarasekare 2002; Bascompte & Jordano 2007; Brooker et al. 2008; Gross 2008; Lankau & Strauss 2007; Roy 2009; San Emeterio, Damgaard & Canals 2007; Vance 1984). The underlying question of the interplay of interspecific interactions is the conception of communities (or ecosystems if abiotic components are taken into account). Historically, the emphasis has been put on trophic links leading to food web as sufficient to characterise community or ecosystem functioning (Goudard & Loreau 2008). Other types of ecological interactions such as interference and mutualism have been often ignored (Arditi, Michalski & Hirzel 2005; Goudard & Loreau 2008). Recent theoretical analyses have shown that communities and ecosystems properties may be strongly modified when non-trophic interactions are included in models. Arditi, Michalski & Hirzel (2005) have shown that communities become ‘super-efficient’ with a certain proportion of mutualistic interactions. In a comparison between interaction and food webs, Goudard & Loreau (2008) showed that the addition of non-trophic interactions change several ecosystem properties such as the shape of the positive relationship between diversity and biomass.

However, the interplay of mutualism, facilitation or interference interactions with exploitation competition has received little attention in view of their potential effects on community dynamics. Coexistence requires negative mutual feedbacks on species abundances to limit interspecific effects with regard to intraspecifics effects. However, mutualism produces positive feedbacks that may prevent coexistence (Schmitt & Holbrook 2003). Few studies explored how mutualism affects coexistence between competitors (Schmitt & Holbrook 2003; Bever et al. 2010). As for interference, several theoretical studies showed that two species competing for the same resource may coexist if interference limits them (Amarasekare 2002; Case & Gilpin 1974; Vance 1984). Interference is in these cases an alternative strategy to increasing competitive abilities (Case & Gilpin 1974). Theoretical models predict that interference is beneficial if the trade-off between its cost and its effect is negative and the resource overlap with the competitor that suffers from interference is high (Case & Gilpin 1974). To ensure coexistence, costly interference such as territoriality or allelopathy may be balanced with beneficial interference such as predation or parasitism that increases the agressor’s growth rate (Amarasekare 2002).

The aim of our work is to experimentally explore interactions between competition, interference and mutualism using protists as model organisms. Laboratory experiments using protists played a key role in the development of population ecology starting with Gause (1935) and continue to play a valuable role in bringing together theory, observations and experiments (Holyoak & Lawler 2005). Indeed, community ecology was kick-started by a set of laboratory experiments conducted by Gause (Gause 1934, 1935; Holyoak & Lawler 2005). As stated by Gause (1935), the models proposed by Volterra (1926) and Lotka (1932) were ‘based on observations of nature’ generating ‘conjectures’. Therefore, the aim of Gause (1935) was ‘to provide a basis with several generations data in order to understand the reciprocal numerical action of animal species living together’ (our translations from French). In agreement with theoretical predictions, Gause showed competitive exclusion with Paramecium aurelia and Paramecium caudatum (Gause 1935). Later, Gill (1972), who reanalysed Gause’s classical competition experiments with P. aurelia, suggested that interference, thanks to noxious endosymbionts, may play an important role in protozoan communities. In a second set of experiments, Gause (1935) showed that Paramecium bursaria can coexist with P. aurelia or P. caudatum. He had chosen species with a priori different ecological niches, P. bursaria feeding predominantly on yeast cells and P. aurelia or P. caudatum on bacterial cells. However, he also conducted experiments with different initial conditions leading to different final equilibria characterised by P. bursaria dominance (Arthur 1987; Gause 1935), which is typical of allelopathic effects (Chao & Levin 1981; Durrett & Levin 1997; Hulot & Huisman 2004). Gause (1935) concluded that these particular equilibria were because of allelochemicals (‘metabolism’s products’) produced by P. bursaria.

In this respect, it may be significant that P. bursaria hosts hundreds of cells of the green algae Chlorella vulgaris and may benefit from carbohydrates synthesised by its endosymbiont presumably in exchange of protection (Karakashian 1963; Pado 1965; Weis 1969). Although the interaction between P. bursaria and C. vulgaris looks mutualistic, it has been suggested that P. bursaria is actually enslaved by its partner (McPhearson 2004). We, therefore, use the neutral term endosymbiosis (Douglas & Smith 1989). Chlorella vulgaris is known to produce allelochemicals that affect the growth of other algal species (Inderjit & Dakshini 1994) and may affect the dynamics of small communities (Hulot, Morin & Loreau 2001). Hence, C. vulgaris may affect P. bursaria’s interaction with other ciliates both as a resource provider and as a producer of allelochemicals.

These findings support the hypothesis formulated by Gause (1935) on the existence of allelopathic effects in his experiments with P. bursaria but suggest also that the endosymbiont could have other roles in ciliate coexistence. Moreover, because the symbiosis of P. bursaria and C. vulgaris is facultative, P. bursaria can be freed of its endosymbiont, which allow studying the effect of this endosymbiosis on the coexistence with other ciliates. Thus, P. bursaria is an interesting model to study the interplay between endosymbiosis, competition and allelopathy. In the following, we will refer to allelopathy as the biochemical interaction, whether stimulatory or inhibitory, among organisms (e.g. Folt & Goldman 1981). This definition generalises Molisch’s (1937) definition of allelopathy that is restricted to primary producers (cited by Gross 2003).

To explore under what conditions endosymbiosis and allelopathy may affect ciliate coexistence, we designed an experiment to separate the effects of resource exploitation, allelopathy and endosymbiosis and to assess their relative importance for the coexistence of P. bursaria with two ciliate species Colpidium striatum and Tetrahymena pyriformis that compete for the same resource, bacteria.

Materials and methods

Strains and culture

The three ciliate species, P. bursaria, C. striatum and T. pyriformis, that were purchased from Carolina Biological Supply© company (Burlington, North Carolina, USA) are bacterivorous. Of these, P. bursaria is the only species to be engaged in a symbiosis: one single P. bursaria individual harbours several hundred green algal cells of C. vulgaris. This endosymbiosis is facultative, and P. bursaria can persist without hosting C. vulgaris (Siegel & Karakashian 1959 cited by Karakashian 1963, 1963). To obtain algae-free paramecia, the herbicide Paraquat (N,N’-dimethyl-4,4′-bipyridinium dichloride) was added to P. bursaria cultures at a concentration of 10 μg mL−1 (Tanaka et al. 2002) and they were cultured at 20 °C under controlled light/dark cycle of 12.12 h. After 2 weeks, there remained some C. vulgaris cells within P. bursaria individuals, but we considered that P. bursaria was effectively free of their endosymbionts (See Fig. S1). We monitored the chlorophyll-a concentration in cultures of endosymbiont-free P. bursaria during pilot experiments. Results showed that this concentration remained very low (data not shown). During all our experiments, including the one in which C. vulgaris was removed from P. bursaria, ciliates were cultured in 50 mL Erlenmeyer flasks containing 15 mL of initially sterile medium (0·75 mg L−1 of powdered Protozoan Pellet in Volvic© water). This medium provides resources for bacteria Serratia marcescens which is consumed by ciliates. Serratia marcescens was the only bacterial species used in the experiments. But despite dilution because of successive culture transplantations, we cannot exclude the presence of other bacterial species coming from purchased protist cultures. Cultures were maintained at 20 °C under controlled light/dark cycle of 12.12 h.

Experimental design

To investigate the importance of endosymbiosis, allelopathy and exploitation competition for the coexistence of P. bursaria with competitors, either C. striatum or T. pyriformis, we conducted two experiments (one for each competitor) with the same experimental design. It comprises seven treatments that were replicated four times: the three species (P. bursaria, endosymbiont-free P. bursaria and a competitor) in monoculture; two competition treatments (P. bursaria or endosymbiont-free P. bursaria with a competitor) and two allelopathy treatments where P. bursaria and its competitor were grown with conditioned medium of the other species. We refer to a monoculture of a given species that is filtered at 0·2 μm to eliminate all individuals and particles and keep only dissolved compounds as a conditioned medium of that species (Fig. 1).These were obtained from monocultures of all species, run in parallel to experimental cultures under the same environmental conditions and with the same initial conditions. These parallel cultures are only a week older than the experimental treatments. Conditioned medium contains organic compounds produced by the protists and bacteria individuals and products of medium degradation. Because conditioned medium does not separate biochemicals produced by the protists from other molecules present in the medium, we refer in the following to conditioned-medium treatment instead of allelopathy treatment.

Figure 1.

 Diagram of the experimental design. Pi, parallel culture of species i; CMi, conditioned medium of species i; Tj, Treatments. i: Paramecium bursaria, endosymbiont-free P. bursaria, Colpidium striatum or Tetrahymena pyriformis. j: monocultures, polycultures and conditioned-medium treatments. Thick grey bar: filtration at 0·2 μm. See text and Table 1 for explanations on the treatments and the CMs added.

Two days before inoculation with ciliates, bacteria S. marcescens were added to the sterile medium in each vial. On day 0, we added to each vial c. 50 cells mL−1 of each ciliate. From day 2, we sampled 1·5 mL of the experimental cultures every 2–3 days of which we used 0·1 mL for protist counting. The remaining 1·4 mL was filtered at 0·2 μm on cellulose acetate filters and kept available for dilution of conditioned medium (Fig. 1). In the seven treatments, replacement of the sampled 1·5 mL follows this design: 0·15 mL of fresh medium with bacteria was added in each vial in addition to a volume of conditioned medium from parallel cultures diluted with the filtered remaining of the sample of the day to reach 1·35 mL (Fig. 1, Table 1). In conditioned-medium treatments, media that were added came from either the competitor or the P. bursaria parallel cultures and were added to reach, in the replacement volume, the ciliate density in the experimental treatment ‘monoculture’ which simulated the effects of a potentially allelopathic species. To control for this addition of conditioned medium, monoculture and competition treatments also received conditioned medium. The volume was calculated to reach, in the replacement volume, the ciliate density in the ‘monoculture’ and ‘competition’ treatments (Table 1). As withdrawals represented 10% of the experimental cultures, the conditioned mediums were added in experimental cultures in proportion of 1 : 10 of species densities. This experimental design induces an impoverishment of culture mediums during the experiment and, as a consequence, population equilibria with lower densities than expected without impoverishment, but it allows to add, on a regular basis, conditioned medium and to separate biochemical effects from competition for resources.

Table 1.   Source of conditioned medium (CM) added in the different treatments and equivalent density in the input. Equivalent density di refers to the density of individuals in treatments before filtration
TreatmentsSource of CM addedEquivalent density di in the input
Monoculture of species iCM of species iDensity of species i in monoculture
Competition of 2 speciesCM of the 2 speciesDensity of the 2 species in competition
Conditioned medium of Paramecium bursaria on the competitorCM of P. bursariaDensity of P. bursaria in monoculture
Conditioned medium of the competitor on P. bursariaCM of the competitorDensity of the competitor in monoculture

For each sample, we counted individuals of each species present in 10 wells of 10 μL. The density in 100 μL was then extrapolated to density in individuals per millilitre. When cells of endosymbiont-free P. bursaria were counted, the colour and the presence of C. vulgaris cells were carefully checked by microscopy to verify that C. vulgaris density in its host remained very low.

Bacterial density measurement at equilibrium and the outcome of exploitation competition

Theoretical models of species competing for one unique resource predict the outcome of exploitation competition in the absence of any other type of interaction. The species that has the lowest resource density at equilibrium when growing alone will exclude its competitors (Grover 1997; Tilman 1982). In the absence of other types of interactions, the resource density at equilibrium, often denoted R*, does only depend on the consumer’s life-history parameters (Grover 1997; Tilman 1982). In addition, it does not take the presence of hosts (virus, endosymbionts) into account. In our system, the R* rule gives a prediction for the outcome of competition between endosymbiont-free P. bursaria and competitors in the absence of any other type of interaction.

Paramecium bursaria hosts an endosymbiont C. vulgaris, which may provide resources (organic carbon) to P. bursaria (Karakashian 1963; Pado 1965; Weis 1969). Here, we will only consider the endosymbiont as an additional resource for P. bursaria (the potential positive effect of P. bursaria on the endosymbiont will not been integrated explicitly). Resource competition theory (Grover 1997; Tilman 1982) summarises also conditions of coexistence of two competitors competing for two resources in the absence of any other type of interaction. Coexistence is ensured if the competitor that is the better exploiter of one resource is the worse exploiter for the other (Phillips 1973) and if each competitor exploits proportionally more of the resource limiting its own growth (León & Tumpson 1975). In our system, C. vulgaris provides resources to P. bursaria that are not exploited by the competitor for bacteria. As a consequence, theoretically, P. bursaria should persist whatever the ability of its competitor to exploit the common resource, bacteria. The outcomes of competition range from coexistence, if the competitor has sufficient abilities to exploit bacteria, to competitive exclusion of the competitor. These predictions describe the expected outcome of exploitation competition between endosymbiont-carrying P. bursaria and the competitors in the absence of any other type of interactions.

We assessed the ability to compete for bacteria of our ciliate species (P. bursaria, endosymbiont-free P. bursaria, C. striatum and T. pyriformis) by measuring in the treatment ‘monoculture’ the bacterial density when ciliates have reached their carrying capacity. To measure bacterial density, we used the Di Aminido Phenyl lndol method (Porter & Feig 1980) on 1 mL of monocultures sampled at equilibrium in each replicate. In the experiment with C. striatum, samplings were carried out on day 46 for P. bursaria and on day 29 for endosymbiont-free P. bursaria and C. striatum. The sampling days are day 32 for P. bursaria and day 28 for endosymbiont-free P. bursaria and T. pyriformis in the experiment with T. pyriformis.

Data analysis

We used R (version 2.8.0) for all statistical analysis. We analysed the effects of treatments on the dynamics of the competitor (C. striatum or T. pyriformis) and P. bursaria. For the experiment with C. striatum, we conducted statistical analysis on subsets of the data only because one or two treatments were missing after day 38 (For data structure, see Fig. 2, Tables S2 and S3). We analysed species densities across time using repeated measure anova, except for C. striatum density at day 40. To perform repeated measure anova, we fitted a linear mixed effect model on species densities using the restricted maximum likelihood method. We tested the effect of time, treatments and time × treatments using a linear mixed effect model where the random effect was the code attributed to each microcosm. For C. striatum density at day 40, we performed an anova using a linear model fitted with generalised least square method. To take into account heteroscedasticity among treatments and over time and improve residuals’ normality, we constructed several models with different variance structure. We selected the best model using the Akaike Information Criterion (See Table S1). We tested the effect of time, treatments and time × treatments on species density using F-tests.

Figure 2.

 Ciliate dynamics in the experiment with Colpidium striatum [Mean ± SE in log (cell mL−1) + 1]. Treatments are represented as follows: monocultures (—), competition (––), and conditioned medium (...). (a) Competition between C. striatum (bsl00066) and Paramecium bursaria (•); (b) Competition between C. striatum (△) and endosymbiont-free P. bursaria (○); (c) C. striatum monoculture (bsl00066) and conditioned medium on C. striatum (bsl00066); (d) P. bursaria (•) and endosymbiont-free P. bursaria (○) monocultures and conditioned medium on P. bursaria (•). (Colour plates with all dynamics are in Fig. S2.)

When treatment effects or time × treatment interactions were significant, we investigated the effects of species interaction and medium, i.e. endosymbiosis, competition and conditioned medium, on the dynamics of the competitor and of P. bursaria using a priori contrast analyses. We compared a priori pooled treatments using orthogonal contrasts. For the analysis of competitor density, the first contrast tested the effect of competition, i.e. the competitor in monoculture or with P. bursaria conditioned medium vs. the competitor with P. bursaria or endosymbiont-free P. bursaria. The second tested the effect of conditioned medium on the competitor, i.e. the competitor in monoculture vs. the competitor with P. bursaria conditioned medium. The third tested the effect of endosymbiosis on competitor density, i.e. the competitor with P. bursaria vs. the competitor with endosymbiont-free P. bursaria. For the analysis of P. bursaria density, the first contrast tested the effect of endosymbiosis, i.e. P. bursaria in monoculture or with competitor conditioned medium or with the competitor vs. endosymbiont-free P. bursaria in monoculture or with the competitor. The second contrast tested the effect of competition for P. bursaria, i.e. P. bursaria in monoculture or with competitor conditioned medium vs. P. bursaria with the competitor. The third contrast tested the effect of competition for endosymbiont-free P. bursaria, i.e. endosymbiont-free P. bursaria in monoculture vs. endosymbiont-free P. bursaria with the competitor. The fourth contrast tested the effect of conditioned medium on P. bursaria, i.e. P. bursaria in monoculture vs. P. bursaria with the competitor. For data after day 38 of the experiment with C. striatum, we modified these contrasts to take into account the absence of one or two treatments (See details in Tables S4 and S6).

The bacterial densities at equilibrium followed normality and homoscedasticity assumptions. For each experiment (with C. striatum or with T. pyriformis), we performed one-way anova followed by Tukey’s post hoc tests to compare the equilibrium bacterial densities of P. bursaria, endosymbiont-free P. bursaria and their competitor.


Experiment with Colpidium striatum as a competitor

Bacterial densities when ciliate monocultures are at their carrying capacity (R*) were significantly different (anova: F2,9 = 94·1, P < 0·0001). Tukey’s post hoc tests showed that the R* is significantly lower for C. striatum than for P. bursaria either with or without its endosymbiont (Table 2). It should be noted that R* is significantly lower for P. bursaria with endosymbionts than it is for P. bursaria without (Table 2). As predicted by R* values, C. striatum did indeed outcompete symbiont-free P. bursaria from day 38 (Fig. 2b). Although bacterial densities at C. striatum carrying capacity are significantly lower than that of P. bursaria, results show coexistence, at least until the end of the experiment (day 49) (Fig. 2a).

Table 2.   Equilibrium value of bacterial density for Paramecium bursaria (inline image), endosymbiont-free P. bursaria (inline image) and their competitor (inline image) Colpidium striatum or Tetrahymena pyriformis. Mean density (SE) of bacteria (cells mL−1) and statistical results in the two experiments. When Tukey’s post hoc tests indicate a significant difference in bacterial densities at equilibrium, the table gives the R* ordering
 Experiment with C. striatumExperiment with T. pyriformis
inline image1·10 × 106 (0·04 × 106)1·49 × 106 (0·20 × 106)
inline image2·40 × 106 (0·60 × 106)4·21 × 106 (0·98 × 106)
inline image0·62 × 106 (0·16 × 106)1·48 × 106 (0·32 × 106)
Tukey’s post hoc testinline image < inline image (P = 0·0134)inline image = inline image (P = 0·9999)
inline image  < inline image (P < 0·0001)inline image < inline image (P = 0·0001)
inline image < inline image (P < 0·0001) inline image < inline image (P = 0·0002)

The analysis of P. bursaria dynamics using repeated measure anovas showed that before day 40, the effect of treatments on P. bursaria density has varied over time (treatments × time: F56,209 = 29·81, P < 0·0001; Table S3) whereas, after day 40, P. bursaria density was affected by treatments only (treatments: F3,12 = 45·51, P < 0·0001; time: F4,46 = 1·19, P < 0·3248; Table S3). Before day 40, the effect of treatments on C. striatum density depends on time (repeated measure anova, treatment × time: F42,168 = 18·40, P < 0·0001; Table S2). On day 40, C. striatum density was affected by treatments (anova: F2,9 = 16·77, P = 0·0009; Table S2). After day 40, repeated measure anova showed C. striatum density was affected neither by treatments (F1,6 = 0·01, P = 0·9042; Table S2) nor by time (F3,18 = 0·49, P = 0·6902; Table S2). Orthogonal contrasts on treatments indicate that endosymbiosis, competition and conditioned medium had a significant effect on the dynamics of the competitor and P. bursaria (contrast analyses are summarised in Table 3 and are detailed in Tables S4 and S6). There was a significant effect of endosymbiosis on P. bursaria densities from the ninth day until the end of the experiment (Table 3, contrast ‘endosymbiosis’). Paramecium bursaria density was significantly higher (Fig. 2a,d; Colour plates with all dynamics are in Fig. S2) than endosymbiont-free P. bursaria density (Fig. 2b,d). Moreover, endosymbiosis has decreased significantly C. striatum density from day 2 to 30 and from day 37 until day 40 (Table 3, contrast ‘endosymbiosis’; Fig. 2a,b). The contrast ‘competition’ indicated that C. striatum and P. bursaria had significant reciprocal effects (Table 3). Paramecium bursaria has a clear negative effect on population density of C. striatum (Fig. 2a,c) and similarly, competition has a negative effect on P. bursaria dynamics (Fig. 2a). Paramecium bursaria densities are significantly lower in the presence of C. striatum (treatment ‘competition’, Fig. 2a) than in its absence (treatments ‘monoculture’ and ‘conditioned medium’, Fig. 2d) at least in the last part of the experiment (from day 21) (Table 3, Fig. S2d). Contrasts on conditioned medium have revealed no significant effects of P. bursaria conditioned medium on C. striatum (Table 3, contrast ‘Allelopathy’; Fig. 2c). But when filtrates of C. striatum cultures were added to P. bursaria, its growth during the exponential growth phase (from 11th to 16th day) was significantly enhanced over P. bursaria growth in monocultures (Table 3, contrast ‘conditioned medium’; Fig. 2d).

Table 3.   Summary of competition, conditioned-medium and endosymbiosis effects on the dynamic of the competitor (Colpidium striatum and Tetrahymena pyriformis) and on Paramecium bursaria (carrying or free of its endosymbiont) in the two experiments. The table summaries days of significant differences (threshold of 5%, NS: non-significant). The upper case letters represent the seven treatments: (A) monoculture of the competitor, (B) competitor with P. bursaria conditioned medium, (C) competitor and P. bursaria in competition, (D) competitor and endosymbiont-free P. bursaria in competition, (E) monoculture of P. bursaria, (F) P. bursaria with competitor conditioned medium and (G)monoculture of endosymbiont-free P. bursaria. Asterisk indicates treatment missing from day 40 (*) and after day 40 (**) in the experiment with C. striatum
 ContrastsExperiment with C. striatumExperiment with T. pyriformis
Effects on competitor dynamic of
 Competition(A*, B**) vs. (C, D)2–11, 16–18, 35–402–4, 11, 21-end
 Conditioned mediumA* vs. B**NS7–9, 25–42, 51–58
 EndosymbiosisC vs. D2–30, 37–402–42, 49-end
Effects on Paramecium bursaria dynamic of
 Endosymbiosis(E, F, C) vs. (G*, D)9-end9-end
 Competition for P. bursaria(E, F) vs. C21–24, 30, 37-end26–42
 Competition for symbiont-free P. bursariaG* vs. D21, 28, 32–3721, 25–30
 Conditioned mediumE vs. F11–1621, 28

Experiment with Tetrahymena pyriformis as a competitor

anova showed significant differences in bacterial density at equilibrium between the three ciliates (F2,9 = 38·2, P < 0·0001). Tukey’s post hoc tests indicate that the R* of endosymbiont-free P. bursaria was significantly higher than the R* of P. bursaria and T. pyriformis (Table 2). No significant difference between the R* of T. pyriformis and of P. bursaria was found (Table 2). As predicted by the R* rule, we observed a competitive exclusion at day 20 of endosymbiont-free P. bursaria by T. pyriformis (Fig. 3b). Because the R* of P. bursaria and T. pyriformis are not significantly different, the prediction of the outcome of exploitation competition is indeterminate. Competition treatments showed that P. bursaria and T. pyriformis coexist during the time course of the experiment (until day 60), but with oscillations of T. pyriformis densities whose oscillation amplitude decreases over time (Fig. 3a).

Figure 3.

 Ciliate dynamics in the experiment with Tetrahymena pyriformis [Mean ± SE in log (cell mL−1) + 1]. Treatments are represented as in Fig. 2. (a) Competition between T. pyriformis (bsl00066) and Paramecium bursaria (•); (b) Competition between T. pyriformis (△) and endosymbiont-free P. bursaria (○); (c) T. pyriformis monoculture (bsl00066) and conditioned medium on T. pyriformis (bsl00066); (d) P. bursaria (•) and endosymbiont-free P. bursaria (○) monocultures and conditioned medium on P. bursaria (•).

Repeated measure anovas of species density across time showed that the interaction treatment × time is significant, for T. pyriformis (F75,299 = 52·02, P < 0·0001; Table S2) and P. bursaria (F84,315 = 53·90, P < 0·0001; Table S3). The contrast analysis showed significant effects of species interactions, i.e. endosymbiosis, competition and conditioned medium, on the dynamics of the competitor and P. bursaria (summarised in Table 3 and detailed in Tables S5 and S7). As with C. striatum, contrasts on endosymbiosis, competition and conditioned-medium effects showed some significant differences (Table 3). Endosymbiosis had a significant effect on P. bursaria dynamics from day 9 until the end and on T. pyriformis from the day 9 to 42 and from day 49 until the end (Table 3, contrasts ‘Endosymbiosis’). Paramecium bursaria density was increased by the presence of endosymbionts (Fig. 3a,b,d), whereas they decreased T. pyriformis density (Fig. 3a,b). Contrasts revealed significant competitive effects on T. pyriformis, P. bursaria and endosymbiont-free P. bursaria (Table 3, contrasts ‘Competition’). Tetrahymena pyriformis in competition with P. bursaria (Fig. 3a) showed first a fast decline: at day 32, its density (10 ± 12 cells mL−1) was much lower than when in monoculture (2465 ± 230 cells mL−1, Fig. 3c) or in competition with endosymbiont-free P. bursaria (1225 ± 151 cells mL−1, Fig. 3b). Then, by the end of the experiment, T. pyriformis density had increased but remained low (30 ± 7 cells mL−1 on day 60, compared to 997 ± 132 cells mL−1 in monoculture) (Colour plates with all dynamics are in Fig. S3). Endosymbiont-free P. bursaria in competition with T. pyriformis became extinct at day 32 whereas it persisted at low densities until the end of the experiment when in monoculture (Fig. 3b,d). In competition with T. pyriformis, P. bursaria densities were significantly lower than in the absence of T. pyriformis from day 16 to 42 (Fig. S3d). The exploration of biochemical-mediated interactions shows that there are no effects, except a negative one on days 21 and 28, of T. pyriformis conditioned medium on P. bursaria dynamics but significant effects of P. bursaria conditioned medium on T. pyriformis dynamics (Table 3, contrasts ‘conditioned medium’). Interestingly, these allelochemical-mediated effects are negative at days 7, 9, 25–42 but then become positive at days 51–58 (Fig. 3c) and match with a small delay the variations of T. pyriformis density in competition treatment (Fig. 3a).


Comparing the dynamics of P. bursaria and C. striatum in competition with their dynamics in monoculture allows us to draw several conclusions. Firstly, according to predictions based on resource competition alone, P. bursaria and C. striatum coexisted until the end of our experiment, while endosymbiont-free P. bursaria went extinct. Endosymbionts can thus mediate the coexistence of the two competitors for the same resource by providing a supplementary resource to its host. Secondly, P. bursaria and C. striatum had reciprocal competitive effects: densities of C. striatum were significantly lower in the presence than in the absence of a competitor either P. bursaria or endosymbiont-free P. bursaria. The same conclusion is reached with the comparison of P. bursaria densities in monoculture and in competition with C. striatum. Thirdly, higher P. bursaria densities when C. striatum conditioned medium was added than in monoculture suggest a transitional facilitative effect of C. striatum medium on P. bursaria because of biochemicals. Therefore, biochemicals may modify the dynamic of species in resource competition. However, we did not observe the decrease in equilibrium densities between the treatment monoculture and the conditioned medium that would indicate that coexistence involves interference between species through allelochemicals (Vance 1984; Case & Gilpin 1974).

In the second experiment, the comparison of ciliate dynamics in monoculture and in competition leads to similar conclusions. In agreement with predictions based on resource competition alone, T. pyriformis outcompeted endosymbiont-free P. bursaria. Tetrahymena pyriformis and P. bursaria coexisted until the end of our experiment, and the low and oscillating densities of T. pyriformis reveal a strong competitive effect of P. bursaria on this ciliate. Thus, competition and monoculture treatments show again a positive effect of the endosymbiont on P. bursaria, which allow the two species to coexist. However, conversely to the experiment with C. striatum, P. bursaria medium affected T. pyriformis until the end of the experiment. These biochemical-mediated effects were first negative to become positive later on. Thus, biochemicals produced in P. bursaria medium elicit different responses of T. pyriformis dynamics ranging from interference to facilitation. Biochemicals effects of P. bursaria medium might have been stronger in the treatment competition than in the treatment allelopathy because biochemicals produced by P. bursaria medium are diluted 10 times in the treatment conditioned medium.

In conclusion, our experiments show that the presence of the endosymbiont C. vulgaris allows P. bursaria to compensate a competitive disadvantage in the exploitation of bacteria. In both experiments, we observed conditioned-medium effects on species dynamics in addition to endosymbiosis and competitive effects, which suggests that biochemical may interact with resource competition.

Endosymbiosis and competition

In the two experiments, comparison of P. bursaria and endosymbiont-free P. bursaria dynamics shows that the presence of C. vulgaris allows higher densities of its hosts. This result seems obvious as we expect an endosymbiont to have a positive effect on its host but deserves some discussion. The green alga is a resource provider for P. bursaria as the ciliate may benefit from products of the photosynthesis (Karakashian 1963; Pado 1965; Weis 1969). Paramecium bursaria also preys upon bacteria. Comparison of bacteria density at equilibrium in P. bursaria and endosymbiont-free P. bursaria monocultures shows that bacteria density is lower in the former than in the latter. In other words, endosymbiont-carrying P. bursaria better exploits bacteria than endosymbiont-free P. bursaria. According to resource competition theory (Grover 1997; Tilman 1982), resource density at equilibrium depends on life-history parameters of the consumer. For instance, if consumption of the resource follows a Monod function and if the consumer has a constant mortality rate, the resource density at equilibrium depends upon the Monod function parameters (uptake rate, affinity constant) and the predator conversion efficiency and mortality rate. Therefore, the decrease in bacteria density at equilibrium in the presence of C. vulgaris suggests that the endosymbiont either increases P. bursaria’s ability to exploit bacteria or decreases its host’s mortality rate. Thus, the endosymbiont C. vulgaris can modify the impact of P. bursaria on bacteria.

Our results show that the presence of the endosymbiont C. vulgaris promotes the coexistence of P. bursaria with its competitors. Resource competition theory (Grover 1997; Tilman 1982) predicts that two competitors may coexist with two resources if, at equilibrium, each competitor is limited by a different resource and if there is a trade-off between competitors’ requirements and their impacts on the resources. If these conditions are not fulfilled, one of the competitors excludes the other. In our system, C. vulgaris provides resources to P. bursaria that are not exploited by the competitor. As a consequence, the endosymbiont provides the opportunity to its host to persist whatever its ability to exploit the common resource.

According to niche theory (see Chase & Leibold 2003 for a history of the niche concept), two species can coexist if they do not share the same niche. In our experiments, the presence of a resource-providing endosymbiont can thus induce niche differentiation and coexistence. When Gause (1935) designed his experiments to explore conditions for coexistence, he manipulated the prey – yeast or bacteria – and space use – bottom or top of the test tub. It is difficult to explain the coexistence that he observed with these two resources as he did not attempt to measure niches (Arthur 1987): neither space use by competitors was not monitored nor prey choice. In addition, there is no data on prey spatial distribution. Nevertheless, here, we show that the presence of C. vulgaris allows P. bursaria’s to differentiate its ecological niche from those of its competitors, which offers a key to understanding coexistence in Gause (1935) experiments. Thus, our results show that endosymbiosis can promote coexistence of competing species by allowing niche differentiation.

The importance of symbionts on the coexistence of species in competition for resources has been observed in plant communities (e.g. Bever et al. 2010). Plants in competition can vary in their dependence on a symbiont (mycorrhizal fungi or symbiotic N-fixer) for a specific soil nutrient (phosphorous or nitrogen, respectively). The coexistence of plants depending on the presence of symbionts may also be interpreted in the light of resource competition theory (Bever et al. 2010; Van der Heijden 2002). A symbiont that increases plant productivity and allows its host to persist at lower levels of a limiting nutrient could contribute to competitive exclusion of other plant species (Hartnett & Wilson 1999; Marler, Zabinski & Callaway 1999) and even reverse the outcome of competition (Hetrick, Wilson & Hartnett 1989). However, these non-shared symbionts can also promote coexistence of plant species. For instance, some plant species are only able to coexist with other plants if mycorrhizal fungi are present, indicating that the symbiont enhances their competitive ability (Grime et al. 1987; Van der Heijden et al. 1998). Similarly, there is strong indirect evidence of positive effect of symbiotic N-fixers on plant coexistence between legumes and non-legumes (Schwinning & Parsons 1996; Cramer, Van Cauter & Bond 2010). Thus, the importance of symbiosis for the outcome of competition is not restricted to our experimental model and seems relevant to other ecological systems. Our results provide a straightforward demonstration of the coexistence mechanism.

Biochemical-mediated interactions

In our experiments, we did observe effects mediated by biochemicals during the transient dynamics. These effects interacted with endosymbiosis differently in the two experiments. In the first experiment, the medium conditioned by C. striatum had a transient positive effect on P. bursaria and may have made the coexistence easier. Indeed, P. bursaria reached its carrying capacity earlier in the treatment competition than in the treatment monoculture. Therefore, this biochemical-mediated facilitation and endosymbiosis had a positive synergetic effect on P. bursaria densities in monoculture. In the second experiment, the comparison between the dynamics of T. pyriformis in monoculture and with P. bursaria conditioned medium reveals direct effects until the end of the experiment of biochemicals present in P. bursaria medium. P. bursaria medium contained biochemicals that first inhibited T. pyriformis growth and then stimulated it. These effects of biochemicals are significant but their strengths (different by an order of magnitude between treatments) are modest. However, in the treatment conditioned medium, biochemicals are diluted up to 1/10 by comparison to biochemicals produced in P. bursaria medium and present in the treatment competition. Thus, significant effects of biochemicals produced in P. bursaria medium suggest that biochemical-mediated interaction may have played a role in coexistence in this experiment. In the treatment competition, T. pyriformis displayed large density changes, which increases its extinction risk, while P. bursaria dynamics was stable. Therefore, we conclude that, in this experiment, endosymbiosis and biochemical-mediated interaction contributed to P. bursaria dominance.

Gause (1935) already suggested that allelopathic interactions occurred in his experiments with P. bursaria. We observed biochemical-mediated interactions, in the experiment with T. pyriformis, where biochemicals produced in P. bursaria medium affected T. pyriformis dynamics from day 25. The absence of conditioned- medium effects at the beginning of the experiment is consistent with the idea that biochemicals have effects when their concentration become higher than a threshold concentration (Chao & Levin 1981; Durrett & Levin 1997; Hulot & Huisman 2004). However, after day 25, the impact of organic compounds produced in P. bursaria medium changed during the time course of the experiment, from negative to positive and matched T. pyriformis density changes in competition treatment. This qualitative change might be the consequence of three hypotheses that are not mutually exclusive.

First, biochemicals could modify demographic rates of T. pyriformis and hence modify the timing of T. pyriformis dynamics. Indeed, the transitory fluctuations displayed by T. pyriformis in the treatment conditioned medium seem to appear earlier than in the treatment monoculture. This timing difference then could correspond to negative followed by positive conditioned-medium effects. Second, the production of biochemicals by P. bursaria could have varied over time. Similar changes in allelochemical-mediated effects had been documented for instance with rye (Reberg-Horton et al. 2005) leading the authors to speak about allelopathy phenology. In our experiment involving animals, the phenology terminology is difficult to use as individuals do not mature as plants. However, the synthesis of biochemicals may have changed directly or indirectly with P. bursaria population dynamics allowing the recovery of T. pyriformis populations in competition and conditioned-medium treatment. Third, the sensitivity of T. pyriformis populations could vary over time because of a selection of T. pyriformis individuals that are less sensitive to biochemicals. Tetrahymena pyriformis population could have been initially mainly sensitive to biochemicals, explaining the observed negative effect. The increase in T. pyriformis densities after reaching very low values could be the result of the selection of individuals insensitive to biochemicals produced in P. bursaria medium. This selection process could occur during a period of 23 days (from day 16 to 39), which amounts to c. 61 generations with a generation time of c. 9 h. Our experiments do not permit more definitive conclusions, but we hope that this discussion about these three hypotheses will stimulate further experimental and theoretical investigations into the mechanisms involved in population response to biochemicals, in particular in the context of competition by exploitation.

Interaction between competition, endosymbiosis and biochemical-mediated interactions

Theory suggests that interference through allelochemicals can promote coexistence (Vance 1984; Case & Gilpin 1974). A weaker competitor for a shared resource can coexist with a stronger competitor if the weaker competitor interferes sufficiently with the other (Vance 1984; Case & Gilpin 1974). In contrast, our results suggest that the coexistence observed in the treatment competition is not caused solely by biochemical-mediated interference. However, in these pioneer models, the resource was not explicit. With the analysis of a model of exploitative and interference competition, Amarasekare (2002) predicted that a species engaged in costly interference such as allelopathy should not be able to coexist with a superior competitor unless it is also engaged in beneficial interference mechanisms such as predation or parasitism. In the experiment with T. pyriformis, where the outcome of competition could not be predicted, P. bursaria coexisted with its competitor, while effects of P. bursaria medium on T. pyriformis did occur. To our knowledge, P. bursaria was not engaged in beneficial interference mechanisms but benefited from the presence of an endosymbiont. Our results suggest that this positive effect was a condition for P. bursaria coexistence with T. pyriformis. In this case, endosymbiosis provided resources analogous to the predation or parasitism hypothesised by Amarasekare (2002) and offered conditions for coexistence. This result may be generalised to plants that cannot develop beneficial interference based on predation or parasitism but are nevertheless engaged in a costly interference competition mediated by allelochemicals. Endosymbiosis is then an additional axis for resource partitioning allowing coexistence.


The results of our two experiments show that endosymbiosis and biochemical-mediated interactions interact and may change the dynamics of populations that are engaged in competition for the same resource. In particular, the presence of an endosymbiont that is a resource-provider ensures the persistence of its host in competition. This might be the key explanation of coexistence in Gause’s (1935) classical experiment. We speculate that a species engaged in competition for a resource and in costly interference such as allelopathy may counterbalance these costs by enlisting a resource-provider endosymbiont instead of embarking on predation or parasitism to harm its competitors.


We thank Minus van Baalen, Sébastien Barot, Jeremy Fox, Michel Loreau and the two anonymous reviewers for helpful comments on the manuscript and Toshinobu Suzaki for his help to obtain chlorella-free P. bursaria. This work was supported by ANR IFB Biodiversité 2005 to F.D. Hulot.