OUP user menu

Spatial variability in the abundance and composition of the free-living bacterioplankton community in the pelagic zone of Lake Bourget (France)

Ursula Dorigo , Dominique Fontvieille , Jean-François Humbert
DOI: http://dx.doi.org/10.1111/j.1574-6941.2006.00139.x 109-119 First published online: 1 October 2006


Spatial variations in the abundance and diversity of the free-living bacterioplankton community of a large Alpine lake, Lake Bourget (France), were investigated in the pelagic zone by means of two two-dimensional samplings taken in 2003. Lake-water samples were collected in winter during water mixing, and in early summer during stratification. The population abundance in each sample was determined by flow cytometry. Denaturing gradient gel electrophoresis of PCR-amplified 16S rRNA gene fragments from organisms measuring less than 2 μm was used to assess eubacterioplankton community composition. In winter, no obvious differences were observed in either the abundance or the diversity of the bacterial community, on either the horizontal or the vertical scales. The only influence detected was that of river water input, but this was at a very minor scale relative to the surface area of the lake. In early summer, changes were found in the community composition on the vertical scale related to the thermal stratification of the water column. There were also marked differences on the horizontal scale at 15 m depth due to internal waves. The implications of these findings for sampling strategies are very important from the perspective of comparative studies of free-living bacterial community diversity and functioning in large and deep lakes.

  • lake
  • bacterioplankton
  • abundance
  • bacterioplankton community composition
  • spatial variability


Since the 1980s, there has been a steady increase in the number of publications about the dynamics and diversity of microbial communities in natural ecosystems (Morris et al., 2002). This subject has attracted considerable research because new molecular tools now make it possible to evaluate these descriptors more effectively, and because there is considerable debate within the scientific community about the origin and functional importance of biodiversity in ecosystems (e.g. Loreau et al., 2002). However, as pointed out in the review by Morris et al., (2002), in the vast majority of these studies, sampling strategies and sample representativity were not clearly defined. Without any data on the repeatability of the evaluation or the intracommunity variability, and by considering only one or two sampling points without knowing whether they are representative of the whole ecosystem studied, it is very difficult to interpret the differences observed between bacterial community compositions when several ecosystems are compared.

Although considerable data are available regarding the spatial variability of microbial diversity in soil ecosystems, investigated at various scales (e.g. Ranjard & Richaume, 2001; Nunan et al., 2002; Ranjard et al., 2003), fewer papers have been publishing that deal with the spatial variability of microbial diversity in aquatic ecosystems and, in particular, in freshwater systems, such as large lakes (Øvreås et al., 1997; Lindström, 1998). Differences in bacterioplankton dynamics and composition may be found at a very small scale as a result of the formation of microscopic or macroscopic organic aggregates (Brachvogel et al., 2001; Knoll et al., 2001). Furthermore, the importance of the influence of the littoral zone or of a river mouth on the composition of the bacterioplankton community is not fully understood, and the impact of allochthonous bacteria on the composition of the bacterioplankton could be very considerable in these areas. Lindström (2001) showed that in a small mesotrophic lake, changes in the bacterioplankton composition could be explained by the massive influx of allochthonous bacteria after heavy rainfall in the nearby catchment area. Similarly, differences in the horizontal or vertical distribution of phytoplankton due to thermal stratification of the water column or to a bloom event, for example, could potentially influence the spatial dynamics and composition of the bacterioplankton community (van Hannen et al., 1999; Gonzáles et al., 2000; Bouvy et al., 2001). So far, little has been discovered regarding the horizontal changes in the distribution of bacterioplankton (Yannarell & Triplett, 2004), whereas more attention has been paid to investigating the distribution along the depth profile (Murray et al., 1998; Dominik & Höfle, 2002; Morris et al., 2002; Humayoun et al., 2003). Palmer et al., (1976) indicated patchiness in the distribution patterns of bacterioplankton based on plate counts both on the small scale (<1 mL) and on the large scale (>1 m horizontal distance), and differences along the depth profile during water stratification. It would appear that horizontal differences are generally less marked than vertical differences, and that the same bacterial assemblages occupy the same depth over large distances along the Catalan coast (Schauer et al., 2000). By contrast, Yannarell & Triplett (2004) investigated the within- and between-lake variability of the composition of the epilimnic bacterioplankton community using automated ribosomal intergenic spacer analysis (ARISA), and found significant horizontal heterogeneity at the scale of hundreds of metres in some lakes.

In the context of our studies comparing the diversity and dynamics of heterotrophic bacterioplankton, and their influence on nutrient cycling in three large subalpine lakes (Annecy, Bourget and Leman) that are characterized by differing trophic levels (ranging from oligotrophic to mesoeutrophic), we chose to investigate the horizontal and vertical distribution of the free-living bacterioplankton community in the pelagic zone of one of these lakes, Lake Bourget. To do this, we performed two-dimensional sampling in winter when there was no thermal stratification in the water column, and in summer when the water column was stratified. The abundance of heterotrophic bacteria in each sample was evaluated by flow cytometry analysis, whereas bacterial diversity was estimated by denaturing gradient gel electrophoresis (DGGE) analysis of a 550 bp fragment of the 16S rRNA gene. Previous research has demonstrated that DGGE is an appropriate way of investigating the temporal and spatial dynamics of major genotypes in aquatic bacterioplankton communities (Murray et al., 1998; Schauer et al., 2000; Yannarell & Triplett, 2004).

Experimental procedures

Site description

Lake Bourget is the largest natural French lake, and is located in eastern France, on the edge of the Alps. It is a warm, mesotrophic, meromictic lake and is elongated (18 and 3 km in length and width, respectively) and orientated north–south. It has a surface area of 42 × 106 m2, a total volume of 3.5 × 109 m3, maximum and average depths of 145 and 80 m, respectively, and a water residence time of approx. 10 years. The lake exhibits summer stratification, and since 1984 the mean water temperature has been 6.4°C, with temperatures of up to 27°C in the first few centimetres during summer. Despite attempts to reduce the phosphorus content (between 1980 and 2001 the level has fallen from 120 to 26 μg L−1), since 1996 blooms of the filamentous and hepatotoxic cyanobacterium Planktothrix rubescens have occurred at least during the summer and autumn periods (Jacquet et al., 2005). A maximum cell concentration of 76 000 cyanobacterial cells mL−1 was found in 1997.

Field sampling strategy

Lake-water samples were collected at the same stations in Lake Bourget during the winter (15 January 2003), when the water mass was mixed, and in early summer (26 June 2003), when the water column was completely stratified. These sampling stations were located across a north–south transect, two east–west transects and in three areas of interest, i.e. in the vicinity of the two main tributaries of the lake (the Leysse and Sierroz rivers, with average flow rates of 8.5 and 3.5 m3 s−1, respectively) and near the outlet (Canal de Savière) of the lake, which is characterized by the presence of reed beds (Fig. 1 and Table 1). On both occasions, samples were collected over a 3 h interval (from 11.30 a.m. to 2.30 p.m.). A sample of 500 mL was collected in bottles, which had previously been autoclaved and rinsed with water from each sample. Sampling was performed at 0 and 15 m at most of the sampling stations. At five sampling stations located on the north–south transect, additional samples were collected at depths of 2, 30 and 50 m. These depths were chosen because they were in the different thermal layers of the water column. In order to estimate the small-scale spatial variability in the distribution of bacterioplankton, in January we collected five additional surface water samples from immediately around our 4-m boat. All the samples were kept at 4°C and in the dark until further processing in the laboratory.


Diagram showing Lake Bourget and the location of the sampling stations along the north–south transect and the two east–west transects. The various sampling stations within areas of major interest, near the river inlets and the lake outlet, are indicated by stars. See Table 1 for the location of the points.

View this table:

Characteristics of the samples used to investigate spatial differences between the study areas. Sampling at 8 m was performed at a point that was too shallow to permit analyses at 15 m

Sample nameDatePositionDepth (m)
SAV115 Jan. 03N45°48,252 E5°49,5900
26 Jun. 030
SAV215 Jan. 03N45°48,151 E5°49,7650, 15
26 Jun. 030, 15
C15 Jan. 03N45°46,527 E5°50,9700, 2, 15, 30, 50
26 Jun. 030, 2, 15, 30, 50
B115 Jan. 03N45°44,776 E5°52,4670, 15
26 Jun. 030, 15
B15 Jan. 03N45°44,800 E5°51,6050, 2, 15, 30, 50
26 Jun. 030, 2, 15, 30, 50
B215 Jan. 03N45°44,826 E5°50,8130, 15
26 Jun. 030, 15
P15 Jan. 03N45°39,606 E5°52,5310, 2, 15, 30, 50
26 Jun. 030, 2, 15, 30, 50
L15 Jan. 03N45°39,369 E5°52,0460
26 Jun. 030
L115 Jan. 03N45°39,401 E5°52,0750
26 Jun. 030
L215 Jan. 03N45°39,421 E5°52,1630, 15
26 Jun. 030, 15
L315 Jan. 03N45°39,454 E5°52,1850, 15
26 Jun. 030, 15
LE26 Jun. 03Mouth of the Leysse River0
Y15 Jan. 03N45°43,533 E5°52,1520, 2, 15, 30, 50
26 Jun. 03
S026 Jun. 03N45°42,088 E5°52,9720
S115 Jan. 03N45°42,089 E5°52,8930, 8
26 Jun. 030, 8
S215 Jan. 03N45°42,108 E5°52,9180, 15
26 Jun. 030, 15
S315 Jan. 03N45°42,104 E5°52,9040, 15
26 Jun. 030, 15
15 Jan. 03N45°42,110 E5°52,2390, 15
26 Jun. 030, 15
S415 Jan. 03N45°42,092 E5°51,8820, 15
26 Jun. 030, 15
A15 Jan. 03N45°41,016 E5°52,1670, 2, 15, 30, 50
26 Jun. 030, 2, 15, 30, 50

Collection of environmental data

At point B (Fig. 1), we used a fluoroprobe (bbe Noldaenke GmbH, Kiel, Germany) to record the vertical profiles of temperature, depth, and in vivo fluorescence, the last allowing us to estimate the chlorophyll a content and major algal classes (Leboulanger et al., 2002). Water subsamples were also taken at point B on 15 January (2, 6, 10, 15, 20, 30 and 50 m) and 26 June (2, 6, 10, 15, 20, 25, 30, 40 and 50 m), and were used to analyse standard nutrients [total organic content (TOC), NH4, NO2, NO3, PO43−, total nitrogen, total phosphorus, particulate phosphorous and SiO2]. Chemical analyses were performed as soon as the samples reached the laboratory, and were carried out in accordance with standard French procedures and protocols (AFNOR, 1982).

Flow cytometry analyses

Subsamples were prepared and used to determine the bacterial abundance. Briefly, 1 mL of the <2 μm fraction was preserved by adding 0.2 μm-filtered glutaraldehyde at a final concentration of 1%. Samples were stored at 4°C for no more than 1 day until being counted using a FACSCalibur (Becton Dickinson) flow cytometer, using the same protocol as described in Dorigo et al., (2004).

Bacterioplankton DNA collection and extraction

Two hundred and fifty millilitres of each lake-water sample was vacuum-filtered through a 2 μm pore-size polycarbonate membrane prefilter (Nucleapore) to eliminate larger eukaryotes (phytoplankton and zooplankton of which the chloroplastidial or mitochondrial 16S rRNA gene would be amplified by the PCR primers used) and not to include P. rubescens cells, which could be present at very high concentrations in Lake Bourget, and whose DNA could overload bacterial community DNA. This prefiltration step also excluded filamentous bacteria other than P. rubescens and particle-associated bacteria. The microbial biomass was then collected and trapped on 0.2 μm pore-size polycarbonate membrane filters (Nucleapore), which were placed in Eppendorf microtubes and to which 750 μL of 55°C prewarmed lysis buffer (40 mM EDTA, 50 mM Tris/HCl, 0.75 M sucrose) was added. The filters were frozen at −80°C until the nucleic acid extractions could be carried out. In January, it took 1 day to filter 50 samples, whereas in June it took 4–5 days to filter 55 samples, due to the higher microorganism content in the summer (see Preliminary experiments below).

Nucleic acid extraction was performed as described in Massana et al., (1997) with some modifications. Briefly, filters in the lysis buffer were placed in a water bath at 55°C for 2 min, vortexed and placed in a sonication bath for 2 min. Lysozyme (Eurobio, 20 000 U mg−1, 2.4 mg mL−1 final concentration) was then added to the filters, and filters were incubated at 37°C for 45 min stirring gently (200 g). Subsequently, SDS (sodium dodecyl sulfate, 1% final concentration) and proteinase K (Eurobio, 30 mU mg−1, 0.2 mg mL−1 final concentration) were added, and the filters were incubated at 55°C for at least 90 min. The lysates were transferred to a fresh Eppendorf tube, and purified twice by sequential extraction with an equal volume of phenol–chloroform–isoamyl alcohol (25 : 24 : 1 by volume, pH 8), and finally an equal volume of chloroform–isoamyl alcohol (24 : 1 v/v). The purified nucleic acids in the aqueous phase were then precipitated by adding 0.1 volume of 3 M sodium acetate (pH 5.2) plus two volumes of cold 99.5% ethanol, stored overnight at −80°C and centrifuged (21 000 g, 30 min). The pellet was washed with 300 μL 80% ethanol, centrifuged again, and then resuspended in 30–100 μL TE (10 mM Tris, 1 mM EDTA, pH 8). The integrity of the total DNA was checked by agarose gel electrophoresis, and quantified by determining the absorbance at 260 nm. Extraction yields were between 38 and 100 μg of DNA per litre of sample in January, and between 35 and 180 μg DNA per litre of sample in June. DNA was stored at −80°C until analysed.

PCR amplification of a 16S rRNA gene fragment

PCR amplifications were performed on 50 μL volumes containing 60 ng of extracted DNA, a 10 ×Taq reaction buffer (Eurobio), 1.5 mM MgCl2, 120 μM of each deoxynucleotide, 1 μM of each primer targeting the 16S rRNA gene (positions 358–907, Escherichia coli numbering system), bovine serum albumin (Sigma, 0.5 mg mL−1 final concentration), and 1.25 U Taq DNA polymerase (Eurobluetaq, Eurobio). Occasionally it was difficult to obtain enough product for DGGE analysis from the PCR reaction. In this case, the amount of DNA was increased from 60 to 120 ng per 50 μL reaction. The primer combination of Eubacteria-specific primer 358f-GC (Muyzer et al., 1993) and universal primer 907rM (Schauer et al., 2003) yielded a DNA fragment of ca. 550 bp. Initially, two distinct thermal cyclers, the Thermal Cycler T-Personal (Biometra) and the PTC-100™ Thermal Cycler (MJ Research Inc.) were used, and the DGGE patterns obtained were compared on a DGGE gel. After comparing the PCR amplification efficiency (higher efficiency for the Biometra cycler) and the DGGE profiles (the same for the two cyclers) of these two methods, all PCR reactions were performed using the Biometra cycler. For each set of reactions, a negative control, in which the template was replaced by an equivalent volume of sterile deionized water, was included. PCR reactions were carried out as described in Schauer et al., (2000). The presence of PCR products of the correct size was confirmed by analysing 4 μL of product on an ethidium-bromide-stained, 1.2% (w/v) agarose gel in 1 × TBE buffer (89 mM Tris-base, 89 mM boric acid, 2 mM EDTA, pH 8.0).


DGGE analysis was performed on PCR fragments, essentially as described in Schauer et al., (2000) by using the CBS-DGGE 2000 system (C.B.S. Scientific, Co. Inc.). The same protocol as described in Dorigo et al., (2004) was used to perform electrophoresis except for the linear gradient of the denaturants urea and formamide, which increased from 40% at the top of the gel to 80% at the bottom. Digital images of the gel were obtained using a Kodak DC290 camera, and were then saved for further analysis using Microsoft Photo Editor software.

DGGE data analysis

The gel images were adjusted for contrast and brightness before analysis. Each DGGE gel was composed of several lanes, each lane corresponding to one sample and consisting of several bands at distinct positions with different intensities. If visual examination was not sufficient, DGGE banding patterns were converted to a binary table using GeneTools software (SynGene, Cambridge, UK) to facilitate comparison of the samples. Two tables were constructed (with samples as rows and DGGE bands as columns): the first took into account the presence or absence of a nucleic acid band at a given height in each lane (scored as 1 or 0, respectively), and the second took into account the relative intensity of each band obtained using the GeneTools software, which allowed us to score the intensity of each band (3, 2, 1 or 0). Both matrices were used to perform cluster analysis; the computing hierarchy based on the Ward method was then performed, using the ADE-4 Software Package (Thioulouse et al., 1997). Pairwise similarity values were calculated using Sorensen's index: Cs=2j/(a+b), where j is the number of common phylotypes (bands) in samples A and B, a is the number of phylotypes in sample A, and b the number of phylotypes in B (Magurran, 1988, p. 179), (Murray et al., 1996). A similarity value of 1 indicates that two DGGE banding patterns are identical, whereas a value of 0 indicates that there are no common bands. Pairwise similarity matrices were constructed to quantify the similarities between DGGE patterns along the depth profile in June, and the similarities between each depth profile in June and the January DGGE profile.

Preliminary experiments

Several preliminary tests were performed to identify the various aspects that had to be taken into account before the Bourget samples were analysed. In general, variability in DGGE fingerprints could be introduced by the amount of water filtered and thus the amount of microbial biomass collected, and by bias during DNA extraction or amplification. Finally, typical forms of bias inherent in the DGGE technique had to be taken into consideration. A first experiment was conducted to establish whether the amount of water filtered influences the DGGE profile; this was determined by filtering progressively decreasing volumes of the same water sample (1000, 500, 250, 110, 70, 50 and 35 mL), and by assessing the DGGE profile after each filter had undergone DNA extraction and PCR amplification. No discrepancies were observed, and we finally decided to filter 250 mL for each water sample analysis.

A second experiment was performed to test the reproducibility of the different steps in the technique. Therefore, in January, five samples were filtered twice on separate filters in order to test the repeatability of the filtration, amplification and DGGE steps. The duplicate filters were processed in parallel (independent filtration, extraction and amplification) and gave identical DGGE patterns on the same DGGE gel.

A third experiment involved storing four samples collected at 2, 15, 30 and 50 m depth, respectively, for a period of 9 days at 4°C, and then analysing them after 0, 1, 3, 6 and 9 days. We were interested in finding out whether the composition of the microflora changes during prolonged storage at 4°C. The DGGE patterns were shown to remain unchanged throughout the 9 days the samples remained in the fridge, whereas the cytofluorometry signature had changed after 6 days to the point of becoming uninterpretable. This experiment was particularly important, because in June filtration was continued over several days. Finally, we checked the degree of reproducibility of replicate samples on the same gel and on different gels. Each microbial assemblage produced a reproducible DGGE fingerprint when run within the same gel; in parallel gels differences were sometimes observed in terms of the presence/absence of minor bands, and sometimes in terms of the migration distance of two bands. These differences were probably related to the quality of the gel. Consequently, as it is difficult to compare samples run in different DGGE gels, two strategies were adopted to compare samples run in different gels. First, at least one or if possible two reference samples were loaded onto each gel each time. Secondly, samples were grouped together in order to make it easier to compare them, e.g. for each sampling date all the surface samples were grouped together on the first gel in order to investigate any horizontal differences at this depth, then all the 15 m samples were run on a second gel and so on. And finally, vertical differences were investigated by running samples taken at different depths at the same sampling station on another gel.


Environmental conditions in Lake Bourget

Figure 2 shows the temperature and chlorophyll a content profiles in winter and in early summer at sampling station B. These data indicate that in winter, the water mass was well mixed at the depths at which samples were collected. Apart from a decrease in chlorophyll a in the upper part of the water column, which was attributable to a device artefact (quenching), which had been identified in another experiment here, the distribution of chlorophyll a was homogeneous throughout the water column (Fig. 2a) with P. rubescens accounting for more than 90% of the total chlorophyll a content (details not shown). Temperature profiles in June revealed a homogeneous epilimnic layer, extending from 0 to 8 m, a metalimnic layer, exhibiting a steep decrease in temperature (between 8 and 18 m) and a hypolimnic layer below 18 m (Fig. 2b). In June, the concentration of chlorophyll a in the upper metalimnion peaked at 20 μg L−1 at a depth of 11 m. At this date, the phytoplankton community was dominated by a green algae (Mougeottia sp.), which accounted for more than 90% of the total chlorophyll a content (details not shown). The data for the chemical analyses for January and June are shown in Fig. 3. In winter, the chemical parameters did not vary with depth (Fig. 3a,b), whereas in summer they did (Fig. 3c,d). The epilimnion showed nitrogen and silicate depletion, and the carbon concentration peaked at a depth of 10 m, and then decreased gradually downwards to the bottom. Phosphorus showed an interesting trend, with a concentration minimum between 20 and 30 m. The phosphate concentrations were low in the top 30 m (with a minimum of 2 μg L−1), indicating that the phytoplankton was most active in this layer. Finally, chemical parameters were homogeneous within each layer, with one exception. In the hypolimnion, phosphorus displayed different concentrations at depths of 30 and 50 m, with values of 2 and 28 μg L–1, respectively.


Vertical profiles of temperature and total chlorophyll a concentration at Bourget station B in January (a) and in June (b).


Vertical distribution of SiO2, total organic content (TOC), NO2, NO3, Ntot, PO4−3 and Ptot in January (a,b), and of SiO2, TOC, NO2, NO3, Ntot, PO4−3, Ppart and Ptot in June (c,d). In January, chemical analyses revealed a homogeneous water mass, apart from the phosphates peak at a depth of 6 m. In June, the conditions were no longer homogeneous (see text).

Flow cytometry counts

In January, taking all the bacterial counts into account, abundances ranged between 0.7 × 106 and 2.2 × 106, with a global average of 1.82 × 106 (±0.28 × 106) cells per mL (Fig. 4). There was no significant difference (Kruskall–Wallis test) between samples with regard to the sampling depth (0, 2, 15, 30 and 50 m) or sampling point. The two lowest bacterial abundances were both observed at sampling points located near the tributaries (L0 and S1).


Average cell counts (± error bars indicating standard deviations) of the nonphotoautotrophic bacterioplankton community of the January samples (white) and of the June samples (striped) in Lake Bourget collected at 0, 2, 15, 30 and 50 m. Counts were made using FCM analysis of SYBR Green I-stained cells.

In June, taking all the bacterial counts into account, abundances ranged between 1.1 × 106 and 12.5 × 106, with a global average of 4.95 × 106 (±2.49 × 106) cells per mL, and there was a significant difference (Kruskall–Wallis test, P<0.05) between samples with regard to depth (0, 2, 15, 30 and 50 m) (Fig. 4). Some significant differences (Mann–Whitney U-test, P<0.05) were observed in the cell counts with regard to both the horizontal and the vertical scales. First, there was a significant increase in the bacterial counts at 30 and 50 m depth compared with those in the upper water layer samples (0 and 2 m) (Mann–Whitney U-test, P<0.05). Bacterial counts in samples taken at 15 m depth were intermediate, but significantly different to counts for the samples from the top and bottom water layers (Mann–Whitney U-test, P<0.05). Secondly, there was a significant difference (Mann–Whitney U-test, P<0.05) in the upper layers (0 and 2 m) between cell counts for the sampling points located away from the edge of the lake, and those in the northern part of the lake (SAV1, SAV2, C, B, B1 and B2) displayed greater abundances than those located in the south (S, S1, S2, S3, S4, A, P, L1, L2, L3). No significant horizontal differences were recorded for the other layers, and cell counts at the sampling points located near the tributaries were no different from those at the other sampling points (Mann–Whitney U-test).

Finally, when comparing bacterial cell counts in January and June, it appears (1) that there was no significant difference between the cell counts at 30 and 50 m depths in June and any of the samples counted in January, and (2) that the variance of the average total cell count at each depth was greater in June than in January.

DGGE fingerprint analyses

In January, 44 of the 49 samples were characterized by having the same DGGE banding pattern (100% homology). We will refer to this DGGE banding pattern as the ‘common January profile’ (all profiles in Fig. 5, except A 50 m). Five samples showed differences when compared with this common January profile, three samples (B1 15 m, Y 30 m and A 50 m) showed differences in terms of the intensity of one, two or three bands, respectively (Fig. 5 for A 50 m, the rest are not shown), whereas two samples, S1 0 m and L 0 m, collected in the vicinity of the two tributaries, were characterized by having one extra band (not shown) and a very different DGGE pattern, respectively. In fact, L 0 m was characterized by having eight common plus six extra bands compared with the common January profile (not shown). Additional samples, which were taken some weeks later (in February near the mouth of the Leysse River) corroborated the DGGE profile of L0 m. Finally, on a small spatial scale, the DGGE pattern of the samples collected around the boat were identical to each other, and to the common January profile.


Negative image of a DGGE gel of PCR amplified 16S rRNA gene fragments (position 358–907) for some January samples. Lanes 1–5: A 0, 2, 15, 30 and 50 m; lanes 6–10: P 0, 2, 15, 30 and 50 m. All the samples except A 50 m (arrows show differences in band intensity) displayed the common January DGGE profile.

In June, a high degree of horizontal homogeneity was observed for the entire lake with regard to samples taken at depths of 0, 2, 30 and 50 m, but not at 15 m. The DGGE profiles were identical for all the samples taken at depths of 0 and 2 m (referred to as ‘the common upper profile’). By contrast, samples taken at 30 and 50 m were characterized by two different DGGE banding patterns. At a depth of 30 m, all the profiles were the same (referred to as ‘the common 30 m profile’), apart from point B, which had five additional bands. At 50 m depth, all the profiles were the same (referred to as ‘the common 50 m profile’), apart from point P, which had one additional band. The cluster analysis applied to the presence/absence of bands within each of the samples from stations C and Y at 0, 2, 15, 30 and 50 m depths and from station B at 0 and 30 m highlights the influence of depth on the bacterioplankton community composition: the upper water layer samples (0 and 2 m) were clustered together, as were the samples taken at 15 m depth, and those taken at 30 and 50 m depth, apart from B 30 m, which differed from the C and Y samples taken at the same depth (Fig. 6).


Cluster analysis of June profiles obtained for sampling stations C and Y at depths of 0, 2, 15, 30 and 50 m, and for B at depths of 0 and 30 m. Samples from the upper water layer (0 and 2 m) were clustered together as were those from the deeper levels.

As already stated, samples taken at 15 m depth showed considerable heterogeneity in DGGE band position and intensity, depending on the sampling point. From the cluster analysis based on the relative intensity of each band, we showed that there was a difference in the community structure depending on the geographical position of the sampling sites. More specifically, along the north–south transect, the more easterly samples were clustered together, as were the more westerly ones. Generally, apart from the 15 m sample from B, which was very different from the other samples taken at this depth, the 15 m DGGE banding patterns were more similar to the 30 m pattern than to the patterns observed at 0 and 2 m depth.

Table 2 shows a matrix used to analyse the pairwise similarities (Cs) between the depth layers of a given station (here P) in June and a representative of the January samples (here C 0 m). The cluster analysis (Fig. 6) also suggests that the similarity between June samples was greatest for samples from 0 and 2 m depth, and that samples collected at 30 and 50 m depth were more similar to those collected at 15 m depth than to those collected at 0 and 2 m depths. The highest and the lowest similarity values between any two June samples were 1 and 0.63, respectively, indicating that the communities were not very different. Table 2 also shows that the January assemblage was similar to that found in June at depths from 0 to 50 m, with Cs values ranging from 0.70 to 0.88. When pairwise similarities were calculated for the January and June samples, it emerged that the January DGGE profile showed the greatest similarity with the June profile for the samples taken at 15 or 30 m depths. Cluster analyses of the fingerprints of this gel corroborate these findings (data not shown).

View this table:

Similarity (Cs) matrix shown as Sorensen's similarity index for bacterioplankton communities collected at 0, 2, 15, 30 and 50 m depth in June and at 0 m depth in January (the common winter profile)

C 0 m JanuaryP 0 m JuneP 2 m JuneP 15 m JuneP 30 m JuneP 50 m June
C 0 m January0.760.760.880.800.70
P 0 m June10.860.660.63
P 2 m June0.860.660.63
P 15 m June0.910.81
P 30 m June0.91
P 50 m June


The aim of this study was to assess the horizontal and vertical variations in the abundance and composition of the free-living bacterioplankton community in the pelagic area of a large alpine lake, Lake Bourget. Sampling was performed at two different times of year (winter and summer) in order to assess the importance of stratification of the water column on these spatial differences.

From 14 to 35 bands were recorded in the DGGE patterns from each sample, which is consistent with previous findings for freshwater bacterial assemblages. Casamayor et al., (2000) reported a total of eight to 17 bands in Lakes Cisü and Vilar. Similarly, Lindström and Bergström (Wintzingerode et al., 1997) found 18–25 bands per sample in two Swedish lakes. Nevertheless, our view of the existing diversity is certainly limited and biased as a result of various pitfalls. It should be kept in mind that DGGE fingerprints reflect the microorganism populations that are present at higher concentrations (Casamayor et al., 2000). These authors report that the number of bands is related to the number of populations that account for more than 0.3–0.4% of the total cell counts. Thus, only numerically dominant species are detected. Muyzer et al., (1993) claim that the reported sensitivity of DGGE is 1% of the template DNA. In addition, there are other problems that may arise during sample collection as a result of insufficient or preferential disruption of cells during the DNA extraction step, or of bias that may arise during the amplification or DGGE steps (gel resolution, gel staining) (Wintzingerode et al., 1997). Therefore, it must be kept in mind that only major changes in the bacterial community composition can be monitored using DGGE.

At the vertical scale, it appeared that in January, the bacterioplankton community of Lake Bourget was homogeneous in terms of both abundance and community composition. Similarly, the physicochemical parameters did not vary with depth. These findings are consistent with mixing of the water column in winter. By contrast, there were marked differences in the abundance and composition of the bacterioplankton community composition in Lake Bourget in June, when the water column was stratified. With regard to abundance, subsurface maximum bacterial values of 6.13 × 106 cells per mL occurred in June, and declined from the peak subsurface chlorophyll level. Morris & Lewis (1992) and Li & Dickie (2001) reported a similar decrease in bacterial abundances from the upper layer to the lower layers. With regard to the composition of the bacterial community, in contrast to the studies of Øvreås et al., (1997) and of Dominik & Höfle (2002), we found that the number of bands increased with depth. Competition between bacterial species for the high level of nutrients in the epilimnic layer could have led to a global increase of bacterial abundance (Fig. 4) and a concomitant decrease in biodiversity. Within the hypolimnic layer, the DGGE banding patterns did depend on the sampling depth (30 or 50 m). It is interesting to note that the phosphorus content was the only chemical parameter measured that differed at 30 and 50 m. As shown in Fig. 3, the total phosphorus and phosphate levels differed at depths of 30 and 50 m. Morris & Lewis (1992) reported that phosphorus may be important in regulating bacterioplankton activity and biomass in aquatic systems, and other studies have highlighted the role of phosphorus in regulating phytoplankton and bacterioplankton growth in aquatic systems (Scanlan & Wilson, 1999; Sala et al., 2002).

On the horizontal scale, no major difference was recorded in terms of either abundance or community composition in January. Only minor differences were observed in the relative intensity of the DGGE bands at three sampling points, suggesting changes in the relative abundance of each species within complex assemblages (Murray et al., 1996). More interestingly, two samples collected near the two tributaries of the lake were characterized by large differences in the bands that were present or absent. These differences were most striking for the sample collected near the outlet of the major tributary (Leysse), which was also characterized by a lower abundance. Recently, Lindström & Bergström (2004) investigated the influence of inlet bacteria on the composition of bacterioplankton in two Swedish lakes characterized by differing hydraulic retention times. They concluded that the input of allochthonous bacteria could be important, depending on the water discharged by the inlet and the retention time of the lake. Consistent with the findings of these authors, in Lake Bourget, which is characterized by having a long retention time (10 years) and a large water volume, the cells imported from tributaries seem to have little impact on the global composition of the bacterioplankton community.

Significant differences were recorded at depths of 0 and 2 m in June in the abundance of bacteria between the sampling points in the north of the lake and those in the south, whereas no differences were observed in the physicochemical parameters or the DGGE banding patterns. During a 2-year survey of the lake, it was also found that there was no obvious difference in the phytoplanktonic community composition at the horizontal scale, but that there were significant differences in the relative abundance of some species (cyanobacteria for example) along a north–south gradient (Freissinet et al., 2004). In the case of cyanobacteria, the influence of some hydrodynamic processes seemed to be implicated in their heterogeneous spatial distribution.

At 30 and 50 m depth, no difference was recorded in the bacterial abundance and community composition. By contrast, considerable heterogeneity in DGGE banding patterns was found in samples collected at a depth of 15 m, which may be related to the hydrodynamic processes (internal waves). More specifically, there was no difference in bacterial cell counts at the scale of the lake, whereas there was a clearly identified east–west gradient in the DGGE banding patterns of samples taken from the north–south transect. On the basis of internal waves, major differences were found on a day-to-day scale in the position of the thermocline in the water column. The depth of the thermocline, and therefore the position of the metalimnic layer, differed by up to 8 m between the north and the south of the lake on the same day of sampling. This was highlighted by monitoring the oxygen concentrations at 13 m depth using an Optode sensor 3930 (Aanderaa, Bergen, Norway). Figure 7 clearly reveals the existence of large internal waves within Lake Bourget on 26 June 2003 (Freissinet et al., 2004).


Variations in temperature (°C) and oxygen concentration (μM) at 13 m depth in Lake Bourget from 25 to 27 June 2003. This figure demonstrates the presence of internal waves that might explain differences in the position of the thermocline at different sampling points on the same day of sampling.

In conclusion, our study revealed a high level of homogeneity in the global composition and abundance of bacterial communities during winter mixing. By contrast, both vertical and horizontal differences were recorded in summer, when the water column was stratified. Different physicochemical, hydrodynamic and biological factors seem to drive these differences. From a practical point of view, we suggest that the vertical temperature profile should be determined in order to make it possible to collect samples from within the layers of interest (the epi-, meta- and hypolimnic layers) when the water column is stratified. Our findings also showed that only a small number of samples may be necessary to provide a reliable and representative estimate of the composition of the bacterial community at the horizontal scale, as long as the samples are obtained from open areas, and well away from inputs, which may affect the structure of the bacterioplankton communities in the lake. Our observations may facilitate future comparisons between large lakes, such as Lakes Bourget, Geneva and Annecy, which are characterized by having different trophic levels and different phytoplanktonic communities.


We would like to thank the ‘Institut de Ciències del Mar’, in particular Carlos Pédros Alio, Laure Guillou and Vanessa Balague for giving U.D. the opportunity to familiarize herself with DGGE. We would also like to thank Stéphan Jacquet for introducing us to flow cytometry, Gérard Paolini and Brigitte Le Berre for their assistance in water sampling and Mathieu Kopec, Manuela Fouqueray and Laurence Volatier for their help with the DGGE analyses. Data on the oxygen concentrations at 13 m depth were provided by Alexis Groleau (LGE, Université de Paris VII). We are grateful for the helpful comments of two anonymous reviewers. Monika Ghosh is acknowledged for revising the English version of the manuscript. This work was partially funded by the Région Rhône-Alpes, Programme Emergence.


  • Editor: Michael Wagner


View Abstract