OUP user menu

Collapse of a Planktothrix agardhii perennial bloom and microcystin dynamics in response to reduced phosphate concentrations in a temperate lake

Arnaud Catherine, Catherine Quiblier, Claude Yéprémian, Patrice Got, Alexis Groleau, Brigitte Vinçon-Leite, Cécile Bernard, Marc Troussellier
DOI: http://dx.doi.org/10.1111/j.1574-6941.2008.00494.x 61-73 First published online: 1 July 2008


Planktothrix agardhii dynamics, microcystin concentration and limnological variables were monitored every 2 weeks for 2 years (2004–2006) in a shallow hypereutrophic artificial lake (BNV, Viry-Châtillon, France). Time-series analysis identified two components in the P. agardhii biomass dynamics: (1) a significant decreasing trend in P. agardhii biomass (65% of the overall variance) and (2) a residual component without significant seasonal periodicity. A path-analysis model was built to determine the main factors controlling the P. agardhii dynamics over the period studied. The model explained 66% of P. agardhii biomass changes. The decreasing trend in P. agardhii biomass was significantly related to a decrease in the PO43− concentration resulting from an improved treatment of the incoming watershed surface water. The residual component was related to zooplankton dynamics (cyclopoid abundances), supporting the hypothesis of a top-down control of P. agardhii, but only when the biomass was low. Forty-nine percent of the variability in the microcystin (MC) concentration (min:<0.1 μg equivalent MC-LR L−1; max: 7.4 μg equivalent MC-LR L−1) could be explained by changes in the P. agardhii biomass. The highest toxin content was observed when P. agardhii biomass was the lowest, which suggests changes in the proportion of microcystin-producing and -nonproducing subpopulations and/or the physiological status of cells.

  • Planktothrix agardhii
  • pico- and nanophytoplankton
  • shallow lake
  • statistical modeling
  • microcystin


Cyanobacterial blooms cause health and environmental concerns (Chorus, 2005). Planktothrix agardhii appears to be one of the most potentially toxic cyanobacterial species in European waters (Nõges & Ott, 2003; Willame, 2005). A recent study of health risks linked to the presence of cyanobacteria in French surface waters (Levi, 2006) showed that P. agardhii was one of the most frequently encountered genera (46% of 1699 samples). Planktothrix agardhii blooms were often associated with the detection of large amounts of microcystin (MC) and toxicity in various biological models (for a review, see Wiegand & Pflugmacher, 2005). The effects of P. agardhii toxic blooms on human health were shown by Codd (1999).

The increase of the occurrence, dominance and bloom of potentially toxic cyanobacteria in the continental waters has been recognized as a direct consequence of the eutrophication of these ecosystems. Over the past 10 years, it has been shown that efforts to reduce eutrophication (oligotrophication) of continental aquatic ecosystems could be successful in reducing noxious cyanobacterial populations, such as P. agardhii (Keto & Tallberg, 2000; Kangro, 2005; Köhler, 2005).

The multifactorial control of P. agardhii dynamics and toxicity has been evidenced by various in vitro studies on P. agardhii strains. They mainly focused on the effects of nutrients (e.g. Ahlgren, 1977; Sivonen, 1990; Znachor, 2006), temperature (Robarts & Zohary, 1987; Sivonen, 1990), light (Sivonen, 1990) and zooplankton (e.g. Davidowicz, 1988; Weithoff & Walz, 1995). While in vitro studies may help to determine the ecophysiological or ecotoxicological characteristics of the strains studied and allow comparing them with other species, only in situ approaches can show the relative effects of multifactorial control on natural populations. However, although a large number of field surveys of P. agardhii have been carried out and limnological characteristics of different lakes have been determined, very few studies used numerical analysis, such as statistical or dynamic models, to detect and quantify the relationships between P. agardhii and environmental variables. The dynamic simulation model of Oscillatoria agardhii in Lake Vechten proposed by Montealegre (1995) appears to be the only example of modeling P. agardhii dynamics in situ. Statistical models were recently shown to be very useful for linking both Microcystis and Anabaena biomass dynamics, and microcystin concentrations to environmental changes in Quebec lakes (Rolland, 2005).

As P. agardhii can produce microcystins and thus cause health concerns, numerical analyses are needed to improve the understanding of the in situ controlling factors of P. agardhii dynamics in different lakes. It is still unclear whether the same factors regulate the biomass of P. agardhii or whether the factors are rather site specific. Finding out more about these factors and their relative importance is a first step in drawing up numerical models able to simulate P. agardhii dynamics and proposing management strategies to reduce the spread of blooms.

The artificial lake at Viry-Châtillon (BNV) located South of Paris is used for leisure activities and was shown to be interesting to further explore both the P. agardhii dynamics and its links with limnological characteristics and microcystin concentrations (Briand, 2002; Yéprémian, 2007). These studies showed (1) the strong dominance of the phytoplankton community by P. agardhii, (2) the associated in situ microcystin concentrations, and (3) the diversity of microcystin variants produced by strains isolated from this ecosystem. A survey was carried out every 2 weeks between 2004 and 2006 to analyze and model the dynamics of P. agardhii in this ecosystem. Conventional time-series analysis and path analysis were used. These statistical analyses are an intermediate approach between dynamics and multiple regression modeling (Legendre & Legendre, 1998). Simultaneous related changes in other phytoplanktonic groups (pico- and nanophytoplankton) were also studied. Finally, P. agardhii dynamics, microcystin concentrations and cellular microcystin content were compared to estimate the usefulness of the P. agardhii biomass as a predictor of microcystin concentration in the BNV ecosystem.

Materials and methods

Study site and sampling procedure

This study was carried out in the ‘Base Nautique de Viry-Châtillon’ (BNV), a water sports center located (2°23′04.2″E, 48°40′03.3″N) in the south suburban area of Paris (France). It is a shallow artificial lake (mean depth: 2.8 m; 98 ha) that receives both diffuse and point source inputs from an 810 ha watershed. The main point source input is a rainfall collector, located on the south shore of the lake collecting untreated surface water from a 430 ha subbasin. In 2003, an in-line treatment plant was added to the rainfall collector to improve the quality of the outflow water released into the BNV. This comprised filtration (3 mm), flotation and decantation to limit the discharge of suspended solids.

The sampling campaign was carried out every 2 weeks between March 23, 2004 and March 21, 2006. Three replicate samples were taken 0.5 m below the water surface at a distance of c. 3 m.

Meteorological variables

Data were obtained for total solar radiation (average between two sampling date) and wind speed at 10-m height (averaged over the 72 h before sampling) from the Orly airport meteorological station (Météo-France), 6 km far from the study site.

Physical and chemical variables

Water temperature, dissolved oxygen and conductivity

Water temperature was measured every 10 min using an in situ thermistor chain (four SEAMON-MINI electronic temperature recorders, accuracy 0.05 °C, Hugrun Inc., Reykjavik, Iceland). The average surface temperature over the period between two sampling dates was then calculated.

Dissolved oxygen and conductivity were measured at each sampling date using a multiparameter Sea-Bird SBE 19 Seacat Profiler (Sea-Bird Electronics Inc., WA).

Chemical analysis

The samples used to quantify dissolved nutrients were immediately filtered using a cellulose acetate 0.22-μm syringe filter (Nalgene, Rochester, NY). Ammonium (NH4+) and orthophosphate (PO43−) analyses were carried out on the day of sampling using a spectrophotometric method as described previously (Greenberg, 1999). The detection limits were 1 μM for NH4+ and 0.1 μM for PO43−. The nitrate (NO3) concentration was measured using a DX600 ion chromatograph (Dionex Corp., Westmont, IL) equipped with an AS14 IonPack analytical column (Dionex Corp.). The detection limit was 0.85 μM.

Raw water was also collected in acid-washed polyethylene containers for total phosphorus (TP) analysis using a spectrophotometric method following acid digestion as described previously (AFNOR, 1982), allowing a detection limit of 0.4 μM.

Biological variables

Chlorophyll a (Chl a) concentration

The Chl a concentration was measured by filtering 300 mL of water (Whatman GF-C filters), followed by methanol extraction (Talling & Driver, 1963) and analysis with a Cary 50 Scan spectrophotometer (Varian Inc., Palo Alto).

Phytoplankton communities

The microphytoplankton species (>20 μm) were determined using samples fixed in buffered formaldehyde (5% v/v for high biomass species) or Lugol (for low biomass species) as described previously (Yéprémian, 2007). The microphytoplankton units were counted using a Malassez counting chamber with a Nikon Optiphot 2 microscope (× 400, Nikon, Melville) or a Utermöhl chamber (Lund, 1951) with an Olympus CK2 inverted microscope (× 400, Olympus Optical Co., Tokyo, Japan) for low phytoplankton biomass. At least 400 phytoplankton units (cells, trichomes, colonies) were counted to reduce the estimation error to <10%. The biovolumes of each species were estimated as described by Sun & Liu (2003). The wet weight biomass was calculated from biovolume estimates, assuming that the phytoplankton cells had a density of 1 g cm−3.

Nano- (2–20 μm) and picophytoplankton (≤2 μm) cells were detected and counted by flow cytometry as described previously (Troussellier, 1993; Campbell, 1994; Crosbie, 2003). Analyses were performed using a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA) with an air-cooled argon laser (488 nm, 15 mW). Subsamples were fixed with buffered formalin (2.5% v/v) and stored immediately in liquid nitrogen until analysis. Cells excited at 488 nm were detected and counted using their right-angle light scattering (RALS) properties and their orange (585/42 nm filter) and red (>650 nm filter) fluorescence from phycoerythrin and chlorophyll pigments, respectively. For each analysis, fluorescent beads (1, 2, 6, 10, 20 μm, Polysciences Inc., Warrington, PA) were systematically added to each sample to standardize the flow cytometer settings. Cell abundances were estimated by adding a known volume of fluorescent beads (True-Counts, Becton Dickinson) with known concentration.

Zooplankton communities

Zooplankton was collected by filtering 1.5 L of water through a 60 μm mesh and then preserved in a 70% ethanol solution (Cottenie, 2003). Abundances (individuals L−1) were estimated using an inverted microscope (Olympus CK2, Olympus Optical Co.) after sample sedimentation. Five major taxonomic groups were distinguished: rotifers, copepod nauplii (NAC), cyclopoid copepods (CYC), calanoids copepods and cladocerans (CLA).

Toxicity assays

The cellular microcystin concentrations of the field samples were estimated using the protein phosphatase 2A inhibition assay (PP2A) as described previously (Briand, 2002). The results are expressed in micrograms equivalent of MC-LR L−1. The detection limit was 0.1 μg equivalent MC-LR L−1 of raw water.

Data analyses

The following successive steps were followed: (1) characterization of the components of P. agardhii biomass dynamics (time-series analyses), (2) identification of the P. agardhii biomass-controlling factors among the different hypothesized ones (path-analysis modeling) and (3) identification of the dynamics components-specific controlling factors (correlation analysis between the identified controlling factors and the components of P. agardhii dynamics).

Time-series analyses were carried out on the 2004–2006 series to identify (1) trends and (2) periodicity, using a linear regression model and contingency periodogram analysis (Legendre, 1981), respectively.

To identify the P. agardhii biomass-controlling factors, a path-analysis model (Wright, 1921) was built on the basis of current knowledge about factors controlling phytoplankton in lakes, including bottom-up, top-down and physical forcing variables (Fig. 1a). The model underlying the path diagram of Fig. 1a was obtained by refining our initial guesses about the potential interrelations between P. agardhii biomass and the other measured variables. For instance, both NH4+ and NO3 have first been introduced as potential N sources controlling P. agardhii biomasses; only NH4+ was significantly related to P. agardhii, and thus retained in the model. Then, in this study, we chose to use PO43− and not TP as an explanatory variable in the path-analysis model. In our case, TP can be considered to be a proxy of phytoplankton biomass as shown by the high correlation coefficient (r=0.91, P<10−4) that can be computed from the dataset. In addition to this, the expected lack of independency between these two variables does not allow including TP as an explanatory variable in a statistical model such as the one used in this study.


(a) Path diagram for the proposed general model. Pa, Planktothrix agardhii biomass (variable to be explained); T, temperature; L, solar radiation; W, wind speed; NH4+, ammonium concentration; PO43−, orthophosphate concentration; NAC, copepod nauplii abundance; CYC, cyclopoid copepod abundance; CLA, cladocerans abundance. (b) Significant direct paths (P≤0.05, arrows) between variable pairs. Only the signs of the significant path coefficients are shown here (values are given in Table 2).

The considered bottom-up factors were the concentration of nutrients (NH4+, PO43−) and light (L). Abundances of NAC, CYC and CLA were considered to be top-down variables. Rotifers were not included in the model as their small size precluded grazing activity on P. agardhii trichomes. Calanoid copepods were not considered due to their scarce abundance (<10 individuals L−1).

Water temperature (T) and wind (W), which act on water column turbulence, were considered to have a potential direct effect on P. agardhii biomass as well as an indirect effect through zooplankton variables.

The Kolmogorov–Smirnov test of goodness of fit, as modified by Lilliefors (1967), was used to test whether the frequency distribution of the variables was normal. When the distribution was not normal, the Box–Cox method (Sokal & Rohlf, 1981) was used to search for the best normalizing transformation. A logarithmic transformation was applied to the NAC, CYC, CLA, NH4+, PO43− and L variables.

The linearity of the relationships between variable pairs was then checked by comparing the significance values in Pearson's r correlation matrices between variables with those found using Kendall's τ matrices. Finally, all variables were normalized (subtracting the mean to each value and dividing by the SD) to remove the effect of the measurement scale of each variable and eliminate the constants in the path analysis equations.

Multicollinearity due to correlation between explanatory variables was checked and, if necessary, taken into account to estimate path coefficients using a ridge regression procedure (Hocking, 1976).

Basic statistics as well as linear or nonlinear models were calculated using statview software (SAS Institute Inc., NC). Kolmogorov–Smirnov and Box–Cox tests were performed using the r package (Casgrain & Legendre, 2004). Path analysis was performed using piste (version 3.1.2), a specific software created by A. Vaudor (Département des sciences biologiques, Université de Montréal).


Limnological characteristics of the BNV

Table 1 presents the mean and range of the main limnological variables recorded during an earlier survey (2001–2002) (Yéprémian, 2007) and during this study (2004–2006) of the BNV.

View this table:

Mean, range, standard deviation and coefficient of variation (%) of the main limnological characteristics, Planktothrix agardhii biomass and % of total biomass, and microcystin concentration as observed during two consecutive studies of the BNV

VariableMeanMinMaxSDCV (%)nMeanMinMaxSDCV (%)n
T (°C)15.24.324.36.945.91113.62.726.17.152.152
DO (mg L−1)10.89.312.81.312.01010.86.614.01.816.736
DO (%)104.488.0125.011.210.71095.874.4133.315.015.729
C (μS cm−1)969.9933.01011.024.12.510938.3627.21176.0167.517.930
NH4+ (μM)
NO3 (μM)8.2<DL35.910.6128.8118.6<DL52.813.5156.950
PO43− (μM)<DL0.70.170.252
TP (μM)14.65.627.87.853.2103.
Secchi depth (m)
Chl a (μg L−1)63.522.0136.531.549.61863.09.9159.632.251.252
Pa (mg L−1)166.381.6605.2157.994.91049.30.0135.142.886.852
Pa (%)98.092.0100.02.92.91080.923.2100.021.426.452
MC (μg equ L−1)9.91.635.08.787.6182.2<DL7.42.093.152
  • T, temperature; DO, dissolved oxygen concentration; DO (%), dissolved oxygen saturation; C, conductivity; TP, total phosphorus; Chl a, chlorophyll a concentration; Pa, P. agardhii biomass; Pa (%), P. agardhii as a percentage of total biomass; MC, microcystin concentration; DL, detection limit.

  • The 2001–2002 series refers to the study carried out by Yéprémian (2007).

The only variables that were significantly different between the two series were PO43− and TP, whose concentrations were lower in 2004–2006 (Table 1). Moreover, PO43− and TP trends, tested by fitting a linear regression between observed values and time, were shown to be significantly decreasing during the 2004–2006 period (PO43−: R2=0.534, P<10−4, Fig. 2c; TP: R2=0.692, P<10−4).


Changes in the main biotic and abiotic variables measured during the 2004–2006 study. (a) Chl a concentration in μg L−1 (Embedded Image), Planktothrix agardhii (○) and Limnothrix redekei (●) biomasses in mg L−1. (b) Picophytoplankton (▲) and nanophytoplankton (⋄) abundances in cells mL−1. (c) NH4+ (♦) and PO43− (◻) concentrations in μM. (d) Cyclopoid abundance (×) in individuals L−1, Temperature (dashed line) in °C.

Vertical temperature profiles did not show any long-term stratification (thermal stratification ≥1 °C m−1 did not last more than 4 days over the period studied; data not shown) over the 2004–2006 period.

Planktothrix agardhii and phytoplankton dynamics

The mean values of P. agardhii biomass decreased from 166 mg L−1 in 2001–2002 to 49 mg L−1 in 2004–2006 (Table 1).

The dynamics of P. agardhii biomasses observed during the 2004–2006 study is shown in Fig. 2a. Following a period of relative stability (from March to December, 2004; mean=97±16 mg L−1), there was a collapse in P. agardhii biomass at the beginning of 2005. After a slight increase during 2006 winter, the biomass decreased again from February 2006. Further ad hoc field observations did not show any return to the high level of P. agardhii biomass (A. Catherine, unpublished data). Such low P. agardhii biomass values were never recorded during the previous study (see the minima values in Table 1).

Chl a concentration variation showed a pattern similar to P. agardhii biomass (Fig. 2a). There was a highly significant linear relationship between Chl a concentration and biomass (R2=0.495, P<10−3). However, only half of the variation in Chl a concentration (Fig. 2a) can be explained by variation in P. agardhii biomass.

Planktothrix agardhii was clearly dominant (>90%) in all the samples of the 2001–2002 series (Table 1). In the 2004–2006 series, the contribution of P. agardhii to the total biomass of microphytoplankton was more variable. At the time of the drastic decrease of P. agardhii (from January to May 2005), other phytoplankton groups increased. In the first part of this period (from January to March 2005), the only other microphytoplankton species detected in significant amount was the filamentous cyanobacteria Limnothrix redekei (Fig. 2a). From April 2005 onwards, this species then showed the same strong decline as P. agardhii. Flow-cytometry analyses outlined that while the microphytoplankton decreased in abundance, there was an increase in nano- and picophytoplankton cell abundance (Fig. 2b): from 8.0 × 101 cells mL−1 for picophytoplankton and 3.3 × 101 cells mL−1 for nanophytoplankton to 105 and 104 cells mL−1, respectively (Fig. 2b). This increase also corresponded to the warm water period in the second year of the survey (Fig. 2d). Abundances of pico- and nanophytoplankton returned to low values at the end of the study period, also corresponding to a decrease of water temperature.

The P. agardhii biomass decreasing trend, tested by linear regression between the observed values and time, was shown to be highly significant (P≤10−3). This linear trend represented almost 65% of the overall variability of P. agardhii (Fig. 3). The linear trend was removed from the observed values to make the residuals stationary. This is a basic condition for testing the periodicity of a time series. The contingency periodogram method was used to search for significant periodicity (P≤0.05) in the residual series. However, no significant period was detected. Thus, while P. agardhii biomass was quite variable during the 2-year survey, neither seasonal nor higher frequency periodic components were detected.


(a) The two components of the Planktothrix agardhii biomass time series: trend (thick line) and residual (○) components. (b) Relationship between P. agardhii (Pa) biomass (in mg L−1) trend values and orthophosphate (PO43−) concentration (in μM). (c) Relationship between P. agardhii (Pa) biomass (in mg L−1) residual values and cyclopoid abundance (in individuals L−1).

Planktothrix agardhii dynamics controlling factors

Figure 1b represents the direct significant paths (P≤0.05) for the variables included in the model (Fig. 1a). Table 2 shows the partition of the total covariance (column A) of each variable pair, X and Y, into direct (XY, column B) and indirect (XZY, i.e. where an indirect pathway between X and Y through Z can be found in the model, column C) and explained and residual (column E) components.

View this table:

Covariance values between variables

Explained covariance (ridge regression)Total explained covariance (multiple regression) (D′)
Bivariate relationships (paths)Total covariance (A)Direct (B)Indirect (C)Total (D=B+C)Residual covariance (E=A−D)Covariance due to multi-collinearity (G=D−D′)
  • The significance of the covariance in columns A and B is:

  • *** P≤0.001,

  • ** P≤0.01,

  • * P≤0.05, or nonsignificant (values in parentheses).

  • The dashes in columns C and E represent covariance excluded by the model design.

    The paths between variables are shown in Fig. 1.

    Pa, P. agardhii biomass; CYC, cyclopoid copepods abundance; NAC, copepod nauplii abundance; CLA, cladocerans abundance; T, temperature; W, average wind speed at 10 m over the 72 h before sampling; L, average global solar radiation between two sampling dates.

The model explained nearly 66% of the P. agardhii changes over the 2 consecutive years of this study.

Among the eight variables that were thought to have a direct action on P. agardhii biomass (Fig. 1a), only three were found to have a significant direct effect (Table 2 and Fig. 1b). The values of path coefficients (Table 2, column B) indicated that PO43− concentration is the most significant variable (0.564, P≤0.001), followed by cyclopoid abundance (−0.355, P≤0.05) and NH4+ concentration (−0.303, P≤0.01). PO43− concentration had a positive influence on P. agardhii biomass. In contrast, both NH4+ concentration and CYC abundance (Fig. 2d) had a negative effect on P. agardhii biomass. The other two types of zooplankton did not show any significant explained covariance with P. agardhii biomass, although their total covariance with P. agardhii was significant (Table 2).

No direct effects of wind (W), light irradiance (L) or temperature (T) were detected (Table 2).

Temperature (T) was found to have a significant effect on the abundance of the three categories of zooplankton (Table 2). This parameter alone explained between 30% and 50% of the variability in zooplankton abundances.

As the P. agardhii dynamics showed a highly significant linear decreasing trend and a higher frequency component, the significant explanatory variables were examined to determine which one could be related to these two different components of P. agardhii dynamics. Correlations were first calculated between P. agardhii biomass values deduced from the linear regression model obtained from the P. agardhii time series and the values of the three significant explanatory variables (CYC, NH4+, PO43−). A highly significant relationship (r=0.825, P≤10−3) was observed only between P. agardhii biomass and Ln PO43− concentrations (Fig. 3b). The decreasing trend observed in P. agardhii biomass during the 2004–2006 study was strongly related to a decrease in the concentration of this inorganic nutrient. Correlations were also calculated between the residual values of the P. agardhii series and the significant explanatory variables. Only CYC abundances were correlated with P. agardhii residual biomass (r=−0.622, P<10−3) (Fig. 3c).

The trend and residual components of the P. agardhii dynamics were, therefore, correlated with different controlling variables. The former was positively correlated with PO43− concentration and the second was negatively correlated with CYC abundance.

Microcystin concentrations

Microcystin concentrations estimated during this study ranged between <0.1 (detection limit) and 7.4 μg equivalent MC-LR L−1 (Fig. 4). These microcystin values were lower than those observed during the previous study (Table 1).


Microcystin concentration (▪) in μg equivalent MC-LR L−1 and Planktothrix agardhii microcystin content (gray bars) in μg equivalent MC-LR mg−1 dynamics during years 2004–2006. *, less than detection limit.

The relationship between the microcystin concentration (y) and P. agardhii biomass (x) was further investigated using a linear regression model. This model (y=0.512+0.034x, R2=0.49, P<10−4) explained a significant part (49%) of microcystin changes. The remaining unexplained variance could result in changes in microcystin content per unit of P. agardhii biomass over the period studied.

Figure 4 shows the P. agardhii microcystin content variations (only for periods when there was a reliable value for the ratio between microcystin concentrations and P. agardhii biomass). Ratios for low microcystin concentrations and/or low P. agardhii biomasses were not calculated (from March to October 2005) as such low values were near the detection limit of the methods used and might produce unrealistically high ratios. During the first period (March 2004 to March 2005), the microcystin content showed a significant increasing trend (R2=0.512, P<10−4) whereas during the October 2005 to March 2006 period, the microcystin content decreased significantly (R2=0.569, P<5.10−3). Planktothrix agardhii biomass showed opposite trends for these two periods (see Fig. 2).

Figure 5 shows the relation between P. agardhii microcystin content and P. agardhii biomass. Planktothrix agardhii microcystin content tended to be higher when P. agardhii biomass was lower. Correlation analysis was unable to detect any significant relationships between the microcystin content and environmental variables.


Relationship between Planktothrix agardhii microcystin content in μg equivalent MC-LR mg−1 and P. agardhii biomass in mg−1during the years 2004–2006.


Environmental characteristics of the BNV

The limnological characteristics recorded clearly show that BNV was a hypereutrophic ecosystem according to the mean and maximum values of standard criteria used for defining the trophic status of continental surface waters (PO43−, TP, Chl a, Secchi depth) (Ryding & Rast, 1994). There was a significant change in the values of some of these indicators between the two surveys (2001–2002 and 2004–2006). Moreover, on splitting the last study into 2 consecutive years (2004–2005 and 2005–2006), the mean values of these indicators exhibited significant (P<0.005) changes (e.g. 2 × for PO43− and Chl a), all indicating an improvement of the trophic status of BNV.

These changes can be attributed, at least to some extent, to the treatment plant that had recently started operation (Bauer, 2003) and induced a reduction of suspended solids (up to 80%, i.e. 330 tons year−1) originating from the main runoff water collector (75% of the total suspended solids getting in the BNV). Moreover, during the period studied, no water from the river Seine, with PO43− concentration higher than the BNV water, entered the lake (B. Tassin, unpublished data).

Perennial vs. nonperennial P. agardhii dynamics

Planktothrix agardhii biomass values observed during the years 2001 and 2002 were in the upper part of the range reported in the literature (e.g. 700 mg L−1 in Lake Albufera; Romo & Miracle, 1993), while other studies gave lower maximum values (e.g. 21 mg L−1 in Lough Neagh; Gibson, 2000).

The measurements of the P. agardhii biomass recorded during the years 2004–2006 in the BNV showed that the dynamics of this cyanobacterial species was different from previous years where biomasses were high and stable (Briand, 2002; Yéprémian, 2007). This study, using statistical analysis, has shown that the main component of P. agardhii dynamics was a significant decreasing trend (65% of the overall biomass variance). The high frequency component of the time series did not show any significant periodicity, at least at the sampling frequency used in this study.

The fast collapse in P. agardhii biomass, declining from 80 mg L−1 to nearly zero in 3 months, was never observed in BNV (since 1998, Briand, 2002). Such a strong decrease and even disappearance of P. agardhii after restoration of a lake or oligotrophication have been reported in other studies (e.g. Chorus & Wesseler, 1988; Keto & Tallberg, 2000; Köhler & Hoeg, 2000; Villena & Romo, 2003; Kangro, 2005; Köhler, 2005).

Relationship between P. agardhii and pico- and nanophytoplankton dynamics

There was a remarkable negative relation between P. agardhii biomass and pico- and nanophytoplankton abundance. According to Mózes (2006), most studies of picophytoplankton dynamics in temperate lakes showed a seasonal trend with highest abundance during the warm water period. The increase of pico- and nanophytoplankton abundance in the warm months (from spring to autumn) of the second year of the survey was not surprising. The maximum abundance of picophytoplankton (105 cells mL−1) was in the middle range of the values reported by Callieri & Stockner (2002) and far from the highest value reported in hypereutrophic lakes (e.g. Lake Apopka, Florida; Carrick & Schelske, 1997). These moderate values of maximum abundance were probably determined by the low concentration of available inorganic nutrients. The fact that the pico- and nanophytoplankton abundance remained low during the warm period of the first year of the survey is more intriguing and might be linked to the high P. agardhii biomass observed over the same period. Such a high biomass may affect (1) the physiology and the growth of picophytoplankton and/or (2) the trophic links of the food chain disturbing the grazing processes of protozoa population such as heterotrophic nanoflagellates and ciliates, the most important grazers of small phytoplankton cells (Callieri & Stockner, 2002).

Factors controlling P. agardhii dynamics

The path analysis model explained almost 66% of P. agardhii variability, a value comparable to the small number of results reported in the literature for other potentially toxic cyanobacteria using multiple regression models (R2=0.71 for Microcystis sp. and R2=0.47 for Anabaena sp.; Rolland, 2005).

Path-analysis modeling showed that P. agardhii dynamics was subjected to both bottom-up and top-down control.

PO43− was the variable showing the highest correlation with the observed changes in P. agardhii biomass. This variable was also highly significantly related to the P. agardhii decreasing trend. The best-fit relationship between PO43− concentrations (x) and P. agardhii biomass trend (y) was not linear (y=44.37 Ln x+130.05, Fig. 5), i.e. P. agardhii biomass did not decrease significantly until PO43− reached concentrations lower than 0.2 μM. This result is in accordance with in vitro and in situ studies showing that the growth rate of P. agardhii decreased when PO43− concentrations were under 0.3 μM (Ahlgren, 1977; Zevenboom, 1982). In BNV, P. agardhii biomass decreased drastically only when PO43− concentrations were below this threshold for at least 3 months (August–December 2004), indicating that the biomass was phosphorus limited. This is consistent with TP concentration (means, minima and maxima reported in Table 1) changes between the 2001–2002 and 2004–2006 periods. As indicated by these values, in 2001–2002, the carrying capacity of this ecosystem did not seem to be limited by phosphorus. The system may, for example, have rather been light limited, whereas in 2004–2006 there will have been a number of occasions for phosphorus limitation. The increase of L. redekei biomass after the decrease of P. agardhii (January–May 2005) may also be linked to the lower phosphorus level as shown by several authors (Rücker, 1997; Wiedner, 2002; Nõges & Ott, 2003).

That NH4+ concentrations were negatively correlated with changes in P. agardhii is probably not due to a direct causal effect. Clearly, the maximum NH4+ values observed cannot be considered to inhibit phytoplankton growth, i.e. to be toxic (Von Rückert & Giani, 2004). This negative correlation could be interpreted as the result of a fast uptake of NH4+ by P. agardhii and/or the result of P. agardhii lysis, leading to NH4+ production by heterotrophic microorganisms through mineralization of decaying algae. A higher NH4+ concentration associated with a lower PO43− concentration may favor the growth of small phytoplankton, which can outcompete, under such circumstances, microphytoplanktonic species such as P. agardhii. Indeed, it is well known that as a source of nitrogen, NH4+ may favor the growth of small phytoplankton species, which are also better adapted to low nutrient concentration due to their larger surface/volume ratio (Koike, 1986; Harrison & Wood, 1988; Cochlan & Harrison, 1991).

The negative relation between cyclopoids and P. agardhii biomass was shown to be associated with the residual variability component of P. agardhii dynamics. The hypothesis of the negative effects of microcystins on zooplankton can be ruled out. The microcystin concentrations (max: 7.4 μg L−1) measured during this study were far below the LD50 reported for various zooplankton organisms (e.g. LD50 expressed in μg MC-LR mL−1: Daphnia sp.=9.6–21.4, De Mott, 1991; Diaptomus bergei=0.45–1, De Mott, 1991; Thamnocephalus platyurus=0.1, Törökne, 1999). It is more probable that the negative correlation can be considered as the result of a top-down control of P. agardhii by cyclopoid grazing. Although adult CYC are generally considered to be primarily omnivorous or carnivorous, it has already been demonstrated that adult females were able to survive and could reproduce on a diet of algae (Hopp, 1997). It has been found that P. agardhii is an edible prey for various zooplankton species belonging to cladocerans (Davidowicz, 1988; Gliwicz & Lampert 1990; Degans & De Meester, 2002) and rotifers (Weithoff & Walz, 1995). Even though, it has been shown that grazing is relatively inefficient when filamentous cyanobacteria reach large biomass (Davidowicz, 1988), the significant decrease in P. agardhii biomass may have allowed zooplankton grazing activity to become more efficient. During the spring and summer of the second year, the large abundance of zooplankton may have helped to maintain P. agardhii biomass at a low level.

The control exerted by the zooplankton on P. agardhii may have been amplified by the decrease in PO43− concentration, limiting the ability of P. agardhii to compensate the loss of biomass due to different processes (e.g. grazing).

Among the environmental variables tested as explanatory variables in the path model, several, especially the temperature, had no significant correlation with P. agardhii biomass. This lack of correlation highlighted that Oscillatoriales such as P. agardhii are able to grow under a wide range of temperatures, even below 10 °C (Post, 1985; Nõges & Ott 2003).

One can expect to improve such a path-analysis model by replacing some of the variables used by other ones more related to biological processes such as grazing, lysis (e.g. viral lysis) and growth rates. However, these variables are more difficult to measure in field studies. The results of the path model can be used to build a predictive equation of P. agardhii biomass using the path coefficients in Table 2, as one can do using multiple regression models (Rolland, 2005). Such predictive equations need to be tested on independent datasets to be validated.

Microcystin concentrations and P. agardhii biomass dynamics

About 49% of changes in microcystin concentration could be explained by P. agardhii biomass. This percentage contrasts with the nonsignificant or low correlation coefficients reported by previous studies in the same ecosystem (Briand, 2002; Yéprémian, 2007) when P. agardhii biomass was more stable. The collapse in P. agardhii biomass during the year 2006 is probably the main explanation for the stronger relationship observed between microcystin concentration and cyanobacterial biomass. Such a collapse extended the range of biomass and microcystin concentration values, giving a better data context to fit a statistical model. Significant relationships between toxin producer biomasses and toxin concentrations have also been reported for studies where a large range of values of these variables has been recorded (e.g. Rolland, 2005; Znachor, 2006; Kardinaal, 2007b).

We also showed that the microcystin content per P. agardhii biomass unit was quite variable and that opposite trends existed between this ratio and biomass. The same relationship has been shown recently in several studies of Microcystis-dominated lakes (Kardinaal & Visser, 2005; Kardinaal, 2007a, b; Welker, 2007). Moreover, using molecular techniques, these studies reported changes in the proportion of either microcystin-producing and -nonproducing genotypes or chemotypes of Microcystis, which could explain the observed changes in microcystin concentration. Such P. agardhii microcystin-producing and -nonproducing subpopulations have already been evidenced in the BNV Lake (Yéprémian, 2007) as well as in other Planktothrix-dominated lakes (Kurmayer, 2004; Welker, 2004).

The increase in microcystin amount per unit of P. agardhii biomass when P. agardhii biomass was low could be the result of an increase of the proportion of microcystin-producing subpopulations in response to variations of environmental parameters. These subpopulations may have specific ecophysiological properties, allowing their better persistence than microcystin-nonproducing subpopulations under adverse conditions. Such differences have been demonstrated for Microcystis under light limitation, showing that microcystin-nonproducing genotypes are stronger competitors than microcystin-producing ones (Kardinaal, 2007a). Thus, it seems that in dense Microcystis blooms where shading plays an important role, microcystin-nonproducing genotypes are superior competitors, leading to seasonal succession. The proportion of these two genotypes in the 2004–2006 P. agardhii series is being estimated using molecular methods, and a study of the ecophysiological properties of several isolated P. agardhii genotypes is being carried out in the BNV Lake (E. Briand, unpublished data).

The microcystin cellular content of the microcystin-producing subpopulations can also be affected by their physiological status. A positive and significant correlation between growth and microcystin-production rates has already been shown for P. agardhii isolated from the BNV Lake (Yéprémian, 2007), P. rubescens (Briand, 2005) and Microcystis (Downing, 2005).

The improvement of predictive models of microcystin concentration in lakes, such as the BNV, requires further investigation of the causes of changes in the microcystin content of toxin producer biomasses. Depending on the type of processes involved (physiological status of the cells and/or subpopulation dynamics), it may be necessary to develop mechanistic models instead of statistical ones.


We are grateful to Aurélie Ledreux for PP2A inhibition assays. The comments of the two anonymous reviewers are greatly appreciated. This work was funded by the ECODYN (INSU no. 04CV131) research program.


  • Editor: Riks Laanbroek


View Abstract