Bacteria can control stoichiometry and nutrient limitation of phytoplankton

Authors

  • M. DANGER,

    Corresponding authorSearch for more papers by this author
  • C. OUMAROU,

    1. Laboratoire Bioemco, Biogéochimie et Ecologie des Milieux Continentaux, UMR 7618 (CNRS, INRA, ENS, Université Paris 6), Ecole Normale Supérieure, 46, rue d’Ulm, 75230 Paris cedex 05, France
    Search for more papers by this author
  • D. BENEST,

    1. Laboratoire Bioemco, Biogéochimie et Ecologie des Milieux Continentaux, UMR 7618 (CNRS, INRA, ENS, Université Paris 6), Ecole Normale Supérieure, 46, rue d’Ulm, 75230 Paris cedex 05, France
    Search for more papers by this author
  • G. LACROIX

    1. Laboratoire Bioemco, Biogéochimie et Ecologie des Milieux Continentaux, UMR 7618 (CNRS, INRA, ENS, Université Paris 6), Ecole Normale Supérieure, 46, rue d’Ulm, 75230 Paris cedex 05, France
    Search for more papers by this author

†Author to whom correspondence should be addressed. E-mail: danger@biologie.ens.fr

Summary

  • 1The chemical composition of primary producers, which occupy the base of food webs, is highly variable over space and time. Determining the causes of these variations is crucial to the understanding of ecosystem functioning.
  • 2By differentially stocking and recycling nutrients, heterotrophic organisms such as bacteria may change the nature of the factor that limits algal growth, thus influencing algal chemical composition.
  • 3We tested this hypothesis by growing a green alga, Scenedesmus obliquus (Turpin) Kützing, along a phosphorus gradient in the presence or absence of a bacterial community.
  • 4We first confirmed that the limiting factor affects the chemical composition of phytoplankton. We then observed that the bacteria switched the nutrient that limited algal growth, from nitrogen to phosphorus in high N : P media. This generated considerable variation in algal stoichiometry.
  • 5We propose that the bacterial compartment may have a greater influence on the structure and functioning of aquatic ecosystems than previously believed.

Introduction

Recently, ecological stoichiometry has emerged as a powerful tool for studying the functioning of both aquatic and terrestrial ecosystems (Elser et al. 1996; Sterner & Elser 2002). Ecological stoichiometry is based on the principle that the transfer of all chemical elements between organisms during ecological interactions is subject to mass-balance constraints (Elser et al. 1996). Primary producers represent the biological level at which carbon is fixed to build living biomass. As primary producers are at the base of trophic food webs, variations in their chemical composition have many consequences for ecological processes (Sterner & Elser 2002). Therefore, understanding the causes of these variations is of key interest when studying food webs and biogeochemical cycles.

Many studies have dealt with the physiological causes of phytoplankton composition, and in particular the close link between algal nutrient content, growth rate (Droop 1973; Rhee 1973; Goldman, MacCarthy & Peavey 1979), and the nature and degree of nutrient limitation (Rhee 1978; Droop 1983; Klausmeier et al. 2004a). Nitrogen and phosphorus are the two main chemical elements studied because they are known to limit primary production in most aquatic and terrestrial ecosystems (Hecky & Kilham 1988; Downing & McCauley 1992). Rhee (1978) demonstrated that algal growth was not limited by N and P simultaneously. Algae were limited either by P or by N on each side of an ‘optimal N : P ratio’. When one element limited algal growth at steady state, it was maintained in constant proportions in algal cells, whereas the other non-limiting nutrient was accumulated in what Rhee (1978) termed ‘luxury consumption’.

In contrast, fewer studies have considered the impact of distinct biological compartments of ecosystems on the stoichiometry of primary producers. Zooplankton grazing and nutrient recycling have been shown to play significant roles in nutrient availability in aquatic ecosystems (Hessen & Andersen 1992; Elser & Urabe 1999). According to stoichiometric principles, the relative quantity of N to P resupplied by herbivores should depend on the difference between their N : P ratios and the N : P ratios of their food (Olsen et al. 1986; Sterner 1990; Urabe 1993; Elser & Hassett 1994). Sterner (1990) hypothesized that a P-rich herbivore (one with a low body N : P ratio) should induce a P limitation in the primary producers it consumes. Such an effect on a nutrient that can limit algal growth should have many consequences for algal chemical composition, and therefore for all the other food-web compartments.

Bacteria represent a compartment of ecosystems that has rarely been considered in previous studies dealing with primary producers’ stoichiometry. Present in all aquatic ecosystems, bacteria play a major role in nutrient recycling. They also represent a potentially important compartment in the total living biomass of aquatic ecosystems (Azam et al. 1983; Cotner & Biddanda 2002). Daufresne & Loreau (2001a) hypothesized that bacteria could generate a switch in the nature of nutrient-limiting algal growth, as do zooplankton (Elser et al. 1988). Thus bacteria potentially could have an important impact on phytoplankton stoichiometry. To our knowledge, such an effect of bacteria on phytoplankton growth and stoichiometry has not been demonstrated previously. In this study, we tested Daufresne & Loreau's (2001a) hypothesis by growing an axenic strain (free of bacteria) of the green alga Scenedesmus obliquus (Turpin) Kützing along a P gradient and in the presence or absence of a natural bacterial community. Based on the results, we propose various potential consequences of the interaction between bacteria and algae for the functioning of aquatic ecosystems.

Materials and methods

experimental design and statistics

Algae were grown along a P gradient in the presence or absence of a natural bacterial community. The experiments were performed in 16 microcosms: four media of distinct N : P molar ratios × two axenicity conditions (presence or absence of bacteria) × two replicates for each treatment. Algae and bacteria were inoculated at time t0, and cultures were maintained for 11 weeks. Cultures were run with a low dilution rate of 0·016 week−1 to allow the biological processes to determine the outcome of the experiments (Olsen, Reinertsen & Vadstein 2002). Each replicate was sampled weekly under axenic conditions to determine algal density and biomass. To determine which nutrient was limiting algal growth within each microcosm at the end of the experiment, cultures were distributed among wells of cellular culture plates to which nutrients were added. For each microcosm, filtrations were carried out at the end of the experiment to determine algal chemical composition. Statistical analyses were performed using superanova software (Abacus Concepts, Inc., Berkeley, CA, USA). When anovas were carried out, Tukey post hoc tests were conducted to determine which pairs of means were significantly different (represented in the figures by different letters). The specific statistical analyses used in each case are described in the relevant sections. For all analyses, we chose a significant probability level of α = 0·05.

biological material

We used an axenic strain of the green alga S. obliquus (CCAP 276/6a) for this investigation. Comparing axenic cultures with cultures to which we added a natural bacterial community (non-axenic treatments) allowed us to determine the effect of the bacteria on algal growth and composition. We assumed that a bacterial community would carry out more ecological functions than a single strain in a particular ecosystem. The bacterial community came from natural lake water (Lake Créteil, France) that was filtered twice through sterile glass microfibre filters (Whatman GF/C, 1·2 µm) to limit contamination by other organisms. Filtered water (1 ml) was added to each non-axenic treatment. It was assumed that this input of water represented a negligible quantity of nutrients compared with the culture media.

culture conditions

Cultures were grown in Erlenmeyer flasks containing 400 ml of a culture medium derived from the standard COMBO medium (1000 µmol N-NaNO3 l−1, 50 µmol P-KH2PO4 l−1, N : P molar ratio 20 : 1, N : P mass ratio 9·03 : 1; Kilham et al. 1998) and used as the N-100%, P-100% reference solution. To create different media compositions along a P gradient, we modified P inputs and held the level of N constant. As the standard COMBO medium was too rich compared with natural water, we diluted its components 10-fold. Therefore P inputs reached 1, 2, 10 and 20% of the COMBO standard P content, while N inputs represented 10% of the COMBO standard N content (providing N : P molar ratios in the basic medium of 200 : 1, 100 : 1, 20 : 1 and 10 : 1, respectively). Erlenmeyer flasks were stopped with silicon porous corks that prevented bacterial contamination but enabled gaseous exchange. All media were then sterilized before algal and bacterial inoculations. Cultures were maintained at a constant 23 °C under 100 µE m−2 s−1 light intensity with a 12-h light : 12-h dark photoperiod. The cultures were stirred using conventional magnetic stirring bars. At the beginning of the experiment, we inoculated each of 16 flasks with algae grown in a standard COMBO medium to a density of 104 cells ml−1. As this inoculum added a small amount of additional nutrients (0·067 µmol P l−1 and 1·44 µmol N l−1) to the experimental medium, we took into account this nutrient supply, and the successive dilutions, in estimating the final N : P ratio of each treatment. The final N : P molar ratios of media for the four treatments were 181 : 1, 95 : 1, 20 : 1 and 10 : 1 (101·12 µmol N l−1 and 0·557, 1·057, 5·056 and 10·055 µmol P l−1, respectively). Note that COMBO medium is initially free of organic C, thus bacteria depend entirely on the algae for dissolved organic carbon (DOC).

sampling and counting algae and bacteria

To determine algal and bacterial densities, four 1·6-ml samples (two for phytoplankton, two for bacteria) were removed weekly under axenic conditions and replaced by sterile medium, at a dilution rate of 0·016 week−1. These samples were fixed with 2% formaldehyde and stored at 4 °C. We verified in preliminary experiments that the optical density of living algal cells at 680 nm (OD680, wavelength of chlorophyll absorption) was significantly correlated with algal biomass (r2 = 0·95, P < 0·0001; unpublished data) using a spectrophotometer (Spectronic 501, Bioblock, Strasbourg, France). Therefore the algal biomass of each culture was determined before fixation by OD680 scans. The algae were then counted at steady state using a photonic microscope (Leitz Diaplan ×30, Wetzlar, Germany) and a Malassez counting grid, which enables the determination of a minimal density of 103 cells ml−1 (Wetzel & Likens 1991). We considered that algal growth had reached a stationary phase when their biomasses did not differ significantly on three successive dates (repeated-measures anova, P = 0·1464). Bacterial densities were also assessed at this stage by counting a minimum of 300 cells per treatment in 4′,6-diamidino-2-phenylindole (DAPI)-stained samples (Porter & Feig 1980) using a microscope fitted for epifluorescence (Leitz Dialux 22 × 1250). At this stage, we also verified the absence of bacterial cells in axenic treatments and the absence of protists in all non-axenic cultures. Note that, even if some small predators of bacteria had been present but not detected in non-axenic media, they would have reduced the bacterial biomass and therefore reduced their competitive effect on algae. Thus the observed effect on algae would be a conservative estimate.

The effect of bacteria and media N : P ratios on estimators of algal biomass at the end of the experiment was determined with two-way anovas (the critical values for bacteria, nutrient, and bacteria–nutrient interaction effects being F1,8, F3,8 and F3,8, respectively). The effect of media N : P ratios on the relative densities of bacteria and algae in non-axenic cultures was determined by one-way anova (critical value for nutrient effect F3,4).

elemental composition of algae

At the end of the experiment, each algal culture was filtered onto preweighed Whatman GF/A glass fibre filters (nominal cut-off 1·6 µm) to separate S. obliquus and free bacterial cells. Thus, for each treatment, we could determine the weight of algal dry matter produced during the experiment. These filters were then used to quantify the percentage of C and N contained in this organic matter using a CHN elementary analyser (NA 1500 Series 2, Fisons, Manchester, UK). Organic P content was determined using the method of Ormaza-Gonzales & Statham (1996). Algal C : N : P ratios are expressed as molar ratios.

The effect of bacteria and media N : P ratios on algal stocks of C, N and P, as well as the N : P, N : C and P : C ratios of S. obliquus at the end of the experiment, were determined with two-way anovas as described above. To assess the differential role of bacteria on these algal parameters according to media composition, we chose to compare the two most different media (N : P = 181 : 1 and 10 : 1) using contrast analysis.

doc concentration

DOC was determined in the axenic algal cultures at the end of the experiment to give an approximate estimate of the relationships between P load, algal exudation rate and C : P-supply ratio for bacterial growth. Total DOC and dissolved inorganic carbon (DIC) concentrations were measured by non-dispersive infrared gas analysis using a Shimadzu TOC-5000 analyser. Total dissolved carbon (TC) was extracted by total combustion at 680 °C, whereas DIC was recovered by acidification with H3PO4. DOC was calculated by subtracting the DIC concentrations from the TC concentrations. The effect of P supply on DOC concentrations was tested by regression analysis.

limitation tests

At the end of the experiment, we assessed which nutrient was limiting algal growth in each microcosm in all treatments. For each culture, the cell suspension was distributed in four wells (5 ml per well). Then 50 µl of each nutrient solution was added to each well: 1·26 mg NaNO3 l−1 (+N), 8·71 mg KH2PO4 l−1 (+P), and all other elements of COMBO medium diluted 10-fold, except N and P (+O). The control (C) received the same volume of pure water. After 5 days’ incubation in conditions similar to the initial cultures, we measured the optical density at 680 nm in each well. We considered the element limiting algal growth to be the one that stimulated an increase in OD680. In the following, we distinguish the terms ‘depleted’ (P-depleted medium = medium with high N : P input ratio) and ‘limited’ (P-limited medium = medium in which algae were stimulated by a P addition).

Statistical analyses on limitation tests were performed separately for each initial N : P ratio and axenicity condition using block anovas without within-block replication. At the end of the experiment, algal densities differed between each of the two replicates of a treatment. As a result of this pre-existing variation in optical density among the Erlenmeyer flasks from which the two series of four wells were established, each Erlenmeyer replicate was considered as a single blocked set of responses to the four addition tests. One degree of freedom being taken by the block effect, the critical value for this limitation test was F3,3. Tukey post hoc tests were conducted to determine which factors limited algal growth.

Results

effects of bacteria on algal biomass along the phosphorus gradient

After inoculation, algal cultures of each treatment showed classic growth patterns. Biomasses first increased exponentially, then growth rates decreased slowly until they reached a stationary phase. At the end of the experiment, the medium composition had a very significant impact on phytoplankton biomass (F3,8 = 11·45, P = 0·003; Fig. 1). Moreover, the significant interaction between P inputs and the presence/absence of the bacterial community (F3,8 = 5·31, P = 0·026) indicated that the significant effect of medium composition on algal biomass in the stationary phase resulted mainly from a significant decrease in algal biomass in the most P-depleted media in the presence of bacteria (Fig. 1). These results were confirmed by contrast analyses between axenic and non-axenic treatments (Table 1). Indicators of algal biomass (OD680, dry weight and algal C l−1) were significantly different in the most P-depleted media (N : P = 181 : 1), whereas there was no significant effect of bacteria in the most P-rich media (N : P = 10 : 1). Moreover, bacteria had a significant negative effect on total P contained in algal biomass at the end of the experiment only in the P-depleted media; total algal N was unaffected by the presence of bacteria (Table 1). In all axenic treatments, total algal C did not differ significantly at the end of the experiment.

Figure 1.

Algal biomass (mg C l−1) at steady state in each medium, in the presence (non-axenic, grey bars) and absence (axenic, open bars) of bacterial communities. The significant interaction (P = 0·02) between phosphorus inputs and the presence of bacteria was largely due to the negative impact of bacteria on algal biomass at steady state in the most P-depleted medium. In the most P-depleted media, bacteria were in competition with algae for P resources. Letters represent groups of treatments significantly different after post hoc Tukey tests (P < 0·05). Vertical lines represent standard errors (n = 2).

Table 1.  Effects of bacteria on indicators of algal biomass (OD680, algal dry matter and algal carbon), and algal nitrogen and phosphorus stocks at the end of the experiments in the most P-depleted medium (N : P = 181 : 1) and in the most P-rich medium (N : P = 10 : 1)
Parameter value at end of experimentN : P = 181 mediaN : P = 10 media
Mean ± SEContrast analysis (P)Mean ± SEContrast analysis (P)
Axenic (N-limited)Non-axenic (P-limited)Axenic (N-limited)Non-axenic (N-limited)
  1. The nutrient limiting algal growth is specified in each case. Means and standard error (n = 2) are shown.

OD6800·064 ± 0·0100·022 ± 0·0010·0010·109 ± 0·0110.094 ± 0.0060·118
Algal dry matter (mg DM l−1) 35·5 ± 6·49 14·6 ± 0·320·00650·87 ± 5·81 57·3 ± 5·70·285
Algal C (mg C l−1) 20·6 ± 4·3  7·4 ± 1·50·008 28·6 ± 4·2 27·8 ± 2·50·849
Algal N (µg N l−1)  904 ± 1  891 ± 970·828  905 ± 16  992 ± 170·181
Algal P (µg P l−1) 14·5 ± 7·2 4·67 ± 0·30·023 94·7 ± 3·5 92·3 ± 0·50·104

algal growth-limiting nutrients

Different patterns were distinguished in the nutrient limitation of algal growth (Fig. 2). During limitation tests in axenic conditions, the OD680 increased significantly with the addition of N (Fig. 2a–d). Thus, in the absence of bacteria, algae were always N-limited in every condition of nutrient availability tested. In contrast, in association with the bacterial community, algal growth was first P-limited in the 95 : 1 and 181 : 1 N : P media (Fig. 2e,f). With an increase in P inputs (10 : 1 and 20 : 1 media), algal growth became N-limited (Fig. 2g,h). Thus, in the two P-depleted media (95 : 1 and 181 : 1), the presence of bacteria induced a switch in the element that limited phytoplankton growth.

Figure 2.

Limitation tests of algae at steady state in the absence (axenic, a–d) and presence (non-axenic, e–h) of the bacteria. Algae were grown in media following a phosphorus gradient, with the level of nitrogen held constant. N : P molar ratios ranged from 181 to 10. The optical density at 680 nm (OD680) was used to represent algal biomass. Limitation tests consisted of adding chemical elements (nitrogen, +N; phosphorus, +P; other essential elements except N and P, +O), then comparing the OD680 of each treatment with the control (C), which received only pure water. Elements limiting algal growth were those that enabled algal growth upon their addition. Letters represent groups of treatments significantly different after post hoc Tukey tests (P < 0·05). Vertical lines represent standard errors (n = 2).

comparison of algal and bacterial densities

Bacterial density was less affected than algal density by nutrient inputs (max/min= 1·29 and 3·10, respectively; Table 2), and the bacteria : algae density ratios were approximately 2·5 times higher in the most P-depleted medium than in the most P-rich medium (Fig. 3; F3,4 = 9·09, P = 0·03). We estimated that bacterial C represented nearly 4% of total C biomass in the most P-depleted medium (Table 2).

Table 2.  Algal and bacterial densities and biomasses at steady state along the phosphorus gradient (mean ± SEM, n = 2)
Medium compositionAlgaeBacteria
Density (×105 cells ml−1)Biomass (mg C l−1)Density (×107 cells ml−1)Estimated biomass (mg C l−1)Bacterial/total biomass (%)
  1. Bacterial biomass estimated from density values with each bacterial cell representing 25 × 10−15 g C cell−1 (Lee & Fuhrman 1987).

N : P = 1812·32 ± 0·767·41 ± 1·451·35 ± 0·170·29 ± 0·063·77
N : P = 953·83 ± 0·5026·1 ± 0·371·45 ± 0·110·36 ± 0·041·36
N : P = 207·20 ± 0·5929·8 ± 1·181·12 ± 0·170·28 ± 0·060·93
N : P = 104·56 ± 0·2427·8 ± 2·531·17 ± 0·090·29 ± 0·031·03
Figure 3.

Relative densities of bacteria compared with algal densities at steady state in non-axenic treatments. Letters represent groups of treatments significantly different after post hoc Tukey tests (P < 0·05). Vertical lines represent standard errors (n = 2).

algal chemical composition

The algal chemical compositions in the four different P treatments followed two patterns (Fig. 4). (i) In N-limited conditions (Fig. 4a and right-hand part of Fig. 4b, nutrient limiting algal growth along P gradient specified under each panel), algal N : C was quite constant regardless of P inputs, whereas P : C ratios increased along the P gradient. (b) In the 95 : 1 and 181 : 1 non-axenic media, when algal growth was P-limited (left-hand part of Fig. 4b), algal P : C ratios remained quite constant, whereas algal N : C ratios decreased with increasing P input. Bacteria had a strong impact on N : C algal ratios (F1,8 = 27·28, P = 0·0008), but this effect was mainly due to the high N : C ratios observed in the two P-limited conditions (Fig. 4b, interaction effect of media N : P ratio × bacteria presence, F3,8 = 23·04, P = 0·0003).

Figure 4.

Evolution of algal chemical compositions in response to a gradient of phosphorus in (a) axenic; (b) non-axenic conditions. Solid lines represent hand-drawn tendencies of P : C molar ratios; dashed lines, N : C tendencies following Rhee's (1978) predictions. Phosphorus inputs are expressed as log(µmol l−1 + 1). Results of the limitation tests are summarized under each graph. Vertical bars represent standard errors (n = 2).

doc concentrations

The final DOC concentrations in axenic cultures reached 4·59 ± 0·04, 4·28 ± 0·19, 3·65 ± 0·19 and 3·64 mg C l−1 for N : P ratios of 181 : 1, 95 : 1, 20 : 1 and 10 : 1, respectively (one sample was lost in the 10 : 1 treatment). We observed a significant negative relationship between DOC concentration and P load: regression analysis, DOC (mg l−1) = −0·106 × P concentration (µmol l−1) + 4·449; r2 = 0·67, n = 7, P = 0·02). Consequently the C : P supply ratio for bacterial growth decreased with the increase in P load.

Discussion

Understanding the factors that determine the stoichiometry of primary producers is of key interest in ecology, particularly when studying food webs and nutrient cycles. Our results confirm that algal chemical composition depends on the nature of the factor that limits their growth, as underlined by Rhee (1978) and theoretically investigated by Klausmeier et al. (2004b) using the same data set. Note that algal compositions measured in our experiment (P : C, N : C and N : P ratios; Fig. 4) varied within the same range of Rhee's values, and were realistic compared with those of previous studies (Rhee 1978; Goldman et al. 1979). In our non-axenic treatments, the increase of algal biomass along the P gradient, in P-limited conditions, induced a decrease in their N content until N became limiting. As expected, the amount of the limiting nutrient in the algae was always minimal and constant in both axenic and non-axenic conditions. Therefore any biotic or abiotic factor effecting this limitation will have an impact on algal stoichiometry.

Our results clearly show that the bacterial component of the aquatic food web has the potential to have a major influence on phytoplankton growth and stoichiometry. In P-depleted conditions, bacteria competed strongly with algae for this element. Despite mesotrophic initial conditions in terms of P load (Wetzel 2001), the very low dilution rate probably induced a rapid immobilization of the majority of P. In these low-P conditions, bacteria have been shown to be better competitors for P than algae (Rhee 1972; Currie & Kalff 1984; Grover 2000). If only competition for the same nutrient had occurred, the bacteria should have excluded the algae. However, bacteria and algae coexisted in our system for several weeks, probably because of the complete dependency of bacteria on the C fixed by the algae. The interaction between plants and decomposers is not only competitive, but also mutualistic (Harte & Kinzig 1993; Daufresne & Loreau 2001b); decomposers allocate sufficient inorganic nutrients for their own growth to maximize their population biomass, leaving the remainder for plant uptake, and in return depend on plant organic C (Harte & Kinzig 1993). The balance between competition and mutualism should depend on C and nutrient supply. In our experiments, the importance of the bacterial compartment was inversely linked to the P content of the medium. Bacteria : algae density ratios were significantly higher in the most P-depleted media. This finding has already been observed in previous studies (for review see Cotner & Biddanda 2002), and in our microcosms was certainly linked to the observed higher DOC exudation by algae with decreasing P availability, as found by others (Bratbak & Thingstad 1985). The bacterial and algal densities observed in our experiment are consistent with their natural range (Cotner & Biddanda 2002).

The major effect of bacteria on the chemical composition of S. obliquus arose from their ability to induce a change in the nature of the nutrient limiting algal growth (Fig. 4). To our knowledge, such an effect has never been shown before. Sterner (1990) have pointed out that P-rich zooplankton, because of the stock of P they represent and their differential recycling of N over P, could induce a P-limitation in their food. In the same way, even though the chemical composition of bacterial communities can be quite variable (Makino & Cotner 2004), bacteria are known to be relatively P-rich (Bratbak 1985; Vadstein & Olsen 1989; Kirchman 2000) and to have a lower average N : P ratio than algae (Vadstein 2000; Cotner 2001; Makino & Cotner 2004). Moreover, bacterial communities often represent a non-negligible biomass in aquatic ecosystems and can reach a significant proportion of the total algal biomass (Azam et al. 1983; Bratbak 1985). Mindl et al. (2005) demonstrated, with the alga Cryptomonas phaseolus and a mixed bacterial assemblage growing in rich P media, that bacterial P represented between 15 and 46% of the total P. In our experiments the bacteria : algae ratio was higher in P-depleted treatments. As discussed previously, this was probably due to the nutrient-induced balance between competition and mutualism. By using a biomass C : P bacterial ratio of 500, equivalent to the maximal bacterial C : P ratio found by Makino & Cotner (2004), we estimated a highly conservative proportion of P immobilized in bacteria in our most P-depleted microcosms, based on their densities. As a maximal value, we considered that the difference in algal P between axenic and non-axenic cultures was due to P immobilization by bacteria. Bacterial P would represent from 4% (C : P = 500) to 67% of available P. In the latter case, the bacterial biomass C : P ratio would be 28, which is realistic when compared with data from previous studies (Makino & Cotner 2004). Thus, in strongly P-limited conditions, bacteria may account for a considerable amount of P in ecosystems and decrease its availability for algae.

The second effect of bacteria on nutrient availability to algae is probably linked to their recycling of organic compounds. In contrast with P, which seems to be easily released from organic matter, most organic N combines strongly with C compounds (Vitousek et al. 2002; Knicker 2004). Thus, in the absence of the action of decomposers or in conditions of low bacterial recycling activity, N might become limiting, even if total N is not negligible. This might explain the appearance of N-limitation in our axenic experimental conditions along the whole P gradient, even with the high N : P ratios of the initial culture medium. In such low-dilution conditions, N was probably rapidly immobilized in the organic matter.

The optimal N : P ratio of a phytoplankton species is often defined on the basis of the medium N : P ratio in which algal growth switches from N- to P-limitation (Rhee & Gotham 1980). This optimal N : P ratio depends on numerous factors. In particular, optimal N : P declines as dilution rate increases, due to a disproportionate increase in P demand with increased growth rate (Sterner & Elser 2002). The very low dilution rate used in our experiment constitutes an additional mechanism explaining N-limitation of algae under axenic conditions. When changing the dilution rate from 0·016 week−1 to 0·20 day−1, we observed a switch from N-limitation towards P-limitation in the most P-depleted axenic cultures (unpublished results), in accordance with the study of Rhee (1978). Note that, despite the low dilution rate, the physiological state of the algae at the end of our experiment was sufficient to allow rapid cell growth in response to the addition of the limiting nutrient.

As a consequence, by differential nutrient immobilization and recycling, bacteria appear to be an important ecosystem compartment that is able to control plant nutrient limitation. In particular, variations in the interactions between algae and bacteria, and modifications of their relative abundances with the seasons or after a change in the quantity of nutrient inputs in ecosystems, may play a major role not only in determining nutrient availability, but also in the identity of the nutrient that limits growth of the primary producers in aquatic ecosystems.

A change in the nature of the limiting nutrient induced by bacteria may have several consequences. Numerous studies have dealt with the nutrient-limitation status of phytoplankton in aquatic ecosystems (e.g. Hecky & Kilham 1988). According to our results, it seems important to take into account the potential role of the bacterial compartment on the identity of the limiting nutrient. Bacteria, with their low N : P ratios and their differential recycling activity, can change the relative quantity of elements available for algae. Consequently, they are able to shift apparent optimal ratios to lower values. The way these optimal N : P ratios are determined should be considered because they are often used in theoretical considerations (for example, for modelling the competition between species of primary producers; Tilman 1982; Grover 1997) or for evolutionary considerations (Klausmeier et al. 2004a).

Within an ecosystemic context and according to stoichiometric principles, changes in factors that limit primary producers’ growth will have many consequences. First, by reducing the nutrient availability for algae, the bacteria should influence phytoplankton community structure. As algal species have distinct optimal nutrient ratios (Rhee & Gotham 1980), variations in nutrient availability due to changes in the relative importance and stoichiometry of the bacterial compartment should also influence the composition of algal communities (Tilman, Kilham & Kilham 1982; Grover 1997; Schulz & Sterner 1999). Furthermore, it has also been shown that cyanobacteria, potentially harmful to the functioning of aquatic ecosystems, are often favoured in N-limited media because of their ability to fix atmospheric N (Smith 1983; Findlay et al. 1994). By reducing P availability, heterotrophic bacterial communities might limit cyanobacterial development in mesotrophic conditions.

Moreover, changes in factors that limit algal growth have important effects on their chemical composition (Fig. 4). Bacteria induced notable changes in algal N : P and N : C ratios; for example, in N : P = 181 : 1 media, algal particulate N : P ratios switched from 71·2 : 1 to 190·1 : 1 and algal N : C ratios increased from 0·04 : 1 to 0·12 : 1 in axenic (Fig. 4a) vs non-axenic media (Fig. 4b), respectively. Algal elemental stoichiometry in part determines ‘algal quality’ for herbivores (Schulz & Sterner 1999). Various studies have demonstrated that food quality has several consequences for the life histories of zooplankton species (Hessen 1992; Sterner et al. 1993; Urabe & Sterner 1996; Park et al. 2003), zooplankton community organization and dynamics (Elser & Urabe 1999; Muller et al. 2001), and secondary production (Sterner et al. 1998). Consequently, a bacteria-induced change in the nutrient content of herbivores’ food may have many consequences for the structure of zooplankton communities. Moreover, bacteria are heavily grazed by the freshwater protist community and are also partially controlled by species such as Daphnia and Diaphanosoma, with the importance of the microbial loop negatively linked to the abundance of such large cladocerans (Christoffersen et al. 1993). The coupling between the structure of the classical food web and microbial processes is complex. Modifications of zooplankton communities induce changes in bacterial composition and activity (Zöllner et al. 2003). Thus the effects of herbivores and bacteria on algal stoichiometry are not necessarily additive. Decomposers and herbivores should be considered together in future models and experimental studies on biotic interactions and ecological stoichiometry.

Our study sheds new light on the need to identify the factors limiting phytoplankton growth in order to understand the ecology of aquatic systems. In these ecosystems, many factors are involved in determining the limiting factor (such as the quantity and stoichiometry of nutrient inputs and outputs). The herbivore compartment, with the stock of nutrients it represents and the differential nutrient recycling it generates, may play a key role in algal nutrient limitation (Sterner 1990). Our results reveal new ways by which the bacterial compartment can influence the structure and functioning of aquatic food webs.

Acknowledgements

This study received financial support from the PNBC (ACI-ECCO) French National Program. We are grateful to T. Daufresne, M. Chérif, F. Darchambeau, X. Lazzaro, F. Maunoury, C. Neill, N. Nunan, and two anonymous reviewers for comments and improvements to the manuscript. We thank E. Aubry and S. Huon for technical help.

Ancillary