OUP user menu

Molecular fingerprinting of lacustrian cyanobacterial communities: regional patterns in summer diversity

Nicolas Touzet, David McCarthy, Gerard T.A. Fleming
DOI: http://dx.doi.org/10.1111/1574-6941.12172 444-457 First published online: 1 December 2013


The assessment of lacustrian water quality is necessary to comply with environmental regulations. At the regional scale, difficulties reside in the selection of representative lakes. Given the risks towards water quality associated with phytoplankton blooms, a mesoscale survey was carried out in Irish lakes to identify patterns in the distribution and diversity of planktonic cyanobacteria. A stratified sampling strategy was carried out via geographic information systems (GIS) analysis of river catchment attributes due to the range of hydrogeomorphological features and the high number of lakes within the study area. 16S rRNA gene denaturing gradient gel electrophoresis analysis showed variation between the cyanobacterial communities sampled, with lower occurrence of cyanobacteria in August concomitant to increased wind and precipitation regimes. Multivariate analysis delineated three ecoregions based on land cover typology and revealed significant patterns in the distribution of cyanobacterial diversity. A majority of filamentous cyanobacteria genotypes occurred in larger lakes contained river catchments with substantial forest cover. In contrast, higher diversity of spherical cyanobacteria genotypes was observed in lakes of lesser trophic state. In the context of aquatic resource management, the combined use of GIS-based sampling strategy and molecular methods offers promising prospects for assessing microbial community structure at varying scales of space and time.

  • cyanobacteria
  • DGGE
  • genetic diversity
  • GIS
  • lakes


The assessment of anthropogenic pressures on aquatic environments is part of globally enforced policies aiming towards the sustainable management of aquatic resources (Bilotta & Brazier, 2008; Viscusi et al., 2008). In Europe, a number of biological variables, including phytoplankton, have been defined as suitable descriptors of water quality through the enactment of the Water Framework Directive (European Communities, 2000). In continental waters, eutrophication has usually been associated with the development of high biomass cyanobacteria blooms (Schindler, 2006; Istvánovics, 2009). Those dominated by toxic species have caused a variety of ecological disruptions and present serious threats towards animal and human health (Codd et al., 2005).

Cyanobacteria are ubiquitous components of phytoplankton assemblages in aquatic environments. They have traditionally been classified according to morphological criteria (Anagnostidis & Komárek, 1988). Phylogenetic studies based on rRNA gene analysis have permitted the assessment of the evolution of different morphotypes and the determination of their lineages, showing that both polyphyletic and monophyletic groups, interspersed in various clades, existed among cyanobacteria (Moore et al., 1998; Willmotte & Herdman, 2001; Moustaka-Gouni et al., 2009). The genetic characterisation of isolates has also enabled the design of taxa-specific molecular markers suitable for the discrimination of morphologically similar species in both culture and environmental samples (Zehr et al., 1997; Rudi et al., 1998; Iteman et al., 2000). To this end, clone libraries and denaturing gradient gel electrophoresis (DGGE) have typically been used for community structure analysis and determining natural microbial succession patterns, including cyanobacteria (Geiss et al., 2004; Song et al., 2005).

The management of aquatic resources and lake habitats necessitates the collection and analysis of representative samples and hence the design of suitable sampling strategies. Studies having investigated the spatial and temporal distributions of lacustrian cyanobacteria over extensive geographical areas encompassing several water bodies are scarce (Catherine et al., 2008; Taranu et al., 2012). Typical limitations have included sampling, methodological, time and budgetary constraints. Advances in geographic information systems (GIS) have permitted the better integration of water bodies with their geomorphologic environment and land usage, which can impact on biodiversity and water quality (Aspinall & Pearson, 2000). As a result, the application of regionalisation frameworks to sampling surveys, combining geographical, geological and hydromorphological data to define contiguous spatial regions that share similar features, has increased in recent years (Detenbeck et al., 2005; Wagner et al., 2007; Cheruveril et al., 2008; Katsiapi et al., 2012).

In the context of lake eutrophication and public health risks incumbent to cyanobacteria, as evidenced from high chlorophyll-a levels and animal deaths in some Irish water bodies (James et al., 1997; Reynolds & Petersen, 2000), a mesoscale survey was carried out in west and north-west Ireland during summer 2009. The survey was undertaken to identify potential aestival and lake trophic state-related patterns in the regional distribution and diversity of planktonic cyanobacteria, which was determined by DGGE analysis of partial 16S rRNA gene. Given the wide range of hydrogeomorphology and land cover type across the region and the high number of water bodies within, a regionalisation framework was applied based on a stratified sampling strategy of lakes according to GIS analysis of river catchment attributes.

Material and methods

Sampling strategy

Sampling was carried out during summer 2009 in a region of Ireland covering a surface of c. 40 000 km2. The study area contained more than 9000 water bodies, accounting for c. 75% of the national amount. Only the lakes larger than 5 ha were considered for sampling, amounting for c. 1200 water bodies, and the sampling effort was set to 50 lakes (c. 4% of population) over a period of 15 weeks (Fig. 2011) via the GIS analysis in ArcMap™ version 9.2 (ESRI®) of topographical and environmental parameters extracted from data sets obtained from the Environmental Protection Agency of Ireland (EPA Ireland) and available online (http://gis.epa.ie/Envision/). Similarity analysis and hierarchical clustering were performed to identify groups of river catchments sharing common traits. The 50 lakes to sample were then proportionally allocated to each of six clusters of river catchments according to the total number of water bodies contained within. Most lakes were selected randomly within the clusters, but five (lakes 04, 08, 15, 33 and 42 in Fig. 2000).

Map of the Republic of Ireland indicating the lakes sampled in the study area during summer 2009 and the delineation of the three ecoregions defined through the analysis of river catchment attributes. The two black circles indicate the locations of the Belmullet (north-west) and Claremorris (central) weather stations.

Lake sampling

Previous work carried out in two western Irish lakes repeatedly showed minor variation in the 16S rRNA gene–based molecular diversity of surface cyanobacteria over a spatial scale of c. 20 km within these two lakes (Touzet N., unpublished data). Based on these results, each selected water body was sampled once during the summer 2009 survey, directly from the shore or using a floating platform. Surface temperature was measured using a handheld instrument, and light penetration in the water column was determined with a Secchi disc, where feasible. The designations sunny spells, overcast and rain were used as meteorological condition descriptors and logged during sampling at each site. Additionally, daily data for global radiation, wind speed and rainfall were obtained from the national meteorological service Met Éireann at the synoptic and weather stations located in Claremorris and Belmullet, respectively (Fig. ). An effort was made to segregate spatially and temporally the sampling occasions within the river catchment clusters so as to encompass potential aestival community succession changes (Table ). Water samples were collected just beneath the surface and preserved with Lugol's iodine in 50-mL tissue culture bottles. Water volumes of 1000 and 100 mL were also filtered onto Whatman GF/F filters (47 mm diameter) and cellulose nitrate membranes (25 mm diameter: 1.0 μm pore size) and stored at −20 °C for chlorophyll-a analysis and nucleic acid extraction, respectively.

View this table:

Information relative to the distribution of the lakes sampled during summer 2009 in west and north-west Ireland

EcoregionRiver catchmentsLakes% of lakes in study areaLakes sampled

Phytoplankton biomass determination and light microscopy analysis

Analysis of chlorophyll-a, used as a proxy for phytoplankton biomass, was carried out by spectrophotometry after overnight extraction of filters in 90% acetone (Lorenzen, 1967). Samples were centrifuged (2880 g, 15 min) prior to carrying out the measurements at 665- and 750-nm wavelengths. Measurements were also taken after adding 0.1 N hydrochloric acid to each extract to compensate for the presence of phaeopigments. The trophic status of the lakes was obtained from EPA Ireland or estimated from a one-point sample via chlorophyll-a concentration and/or Secchi depth according to OECD (1982).

In complement of molecular analysis, preserved water samples were examined by inverted light microscopy after overnight phytoplankton settlement in 50-mL flat-bottomed bottles. Order-level identifications were carried out at × 100–400 magnifications based on morphological descriptions from the literature (John et al., 2002).

Nucleic acid extraction from filtered surface water samples

DNA extraction was carried out using the DNeasy® Blood & Tissue Kit (Qiagen, UK) according to the manufacturer's instructions for Gram-negative bacteria. Minor modifications to the protocol were carried out based on the previous successful recovery of DNA from marine dinoflagellate cysts, notoriously resistant to cell lysis (Touzet N., unpublished data). First, sample-containing membranes were disrupted by bead-beating in a RiboLyser (Hybaid, UK) (full power, three cycles of 20 s each) using 180-μm-diameter glass beads. Cell lysis (56 °C, 180 min) was then carried out with proteinase-K in a Thermomixer® (Eppendorf, UK). DNA was eluted in a final volume of 100 μL with PCR grade water and extract concentration and quality determined using a NanoDrop 1000 (Thermo Scientific, UK). Samples were stored at −20 °C until use.

Cyanobacteria community fingerprinting: semi-nested 16S rRNA gene DGGE PCR

The initial amplification of partial 16S rRNA gene was carried out using the primer combination CYA359f (5′-ggggaattttccgcaatggg-3′) and 23S30r (5′-cttcgcctctgtgtgcctaggt-3′) (Nübel et al., 1997; Taton et al., 2003). Each PCR contained 1× buffer, 2 mM MgCl2, 200 μM dNTPs, 0.25 μM of both primers, 1 unit of GoTaq® polymerase (Promega) and 1 μL of template DNA. The thermocycling conditions were as follows: initial denaturation step (94 °C, 5 min), 15 cycles of amplification [denaturation (94 °C, 0.5 min), annealing (58 °C, 1 min) and extension (72 °C, 2 min)], and a final extension step (72 °C, 10 min). The second master mix had a similar composition and was supplemented with 1 μL of the first PCR prior to subjecting the samples to the second amplification carried out with the primers CYA359f and CYA781r(a) (5′-gactactggggtatctaatcccatt-3′) or CYA781r(b) (5′-gactacaggggtatctaatcccttt-3′) to amplify preferentially two distinct cyanobacterial types, named for the purpose of this study genotype groups I and II (GG-I and GG-II) and elsewhere referred to as filamentous and spherical cyanobacteria, respectively (Boutte et al., 2006). The GC clamp (5′-cgcccgccgcgccccgcgcccgtcccgccgcccccgcccg-3′) was placed at the 5′ end of the forward primer. The cycling conditions were as follows: initial denaturation step (94 °C, 5 min), 35 cycles of amplification [denaturation (94 °C, 1 min), annealing (62 °C, 1 min) and extension (72 °C, 2 min)] and a final extension step (72 °C, 30 min).

Amplicons were visualised by 1% agarose gel electrophoresis after staining with SYBR® Safe (Invitrogen). Separation of individual amplicons was performed by DGGE using the DCode Mutation Detection System (Bio-Rad Laboratories). Gels were constituted of 8% acrylamide/bis (37.5 : 1) in 1×-TAE with a 45–55% denaturant gradient prepared from a 100% denaturation solution of 7 M urea and 40% formamide (v/v). Vertical electrophoretic separation of amplicons was carried out for 16.5 h at 60 °C at a constant voltage of 60 V. Replicate DGGE profiles were run to verify the consistency of band separation for selected lake samples. Some samples were included across several gels as migration controls for comparative purposes. Gels were stained in 300 mL of 10 mg mL−1 ethidium bromide in 1×-TAE buffer for 40 min and washed in 300 mL of deionised water for 12 min. All gels were imaged by UV illumination using a G:Box (Syngene, UK). Analysis of the DGGE migration patterns was carried out using the Total-Lab TL120 software (Nonlinear Dynamics, UK). A densitometric scan of the gels was created, and background noise was subtracted using a rolling disc algorithm. Band matching matrices were constructed using peak height values. The matrices of band intensities for cyanobacteria GG-I and GG-II were analysed by hierarchical clustering.

16S rRNA gene band excision, sequencing and phylogenetic inference

A number of 34 bands of both high and low intensity were selected at random throughout the gels. This corresponds to a c. 15% subsample of the total number of bands detected in the gels and could be considered suitably representative. The bands were excised from the acrylamide gels and eluted overnight at 5 °C in 100 μL of PCR grade water. Re-amplification was then performed with 1 μL of PCR template, the combination of seq-1f (5′-gcgaaagcctgacggagc-3′) and seq-1r (5′-ggggtatctaatcccattcgct-3′) internal primers and the 35-cycle thermocycling programme. Amplicons were purified using the QIAquick PCR Purification Kit (Qiagen) and eluted in 30 μL of PCR grade water prior to external sequencing (Eurofins-MWG, Germany).

Upon reception, the sequences were screened with blastn (Altschul et al., 1990: http://www.ncbi.nlm.nih.gov) to orientate identifications, compiled then with other cyanobacterial sequences imported from GenBank and aligned with the pairwise alignment function of GeneDoc and ClustalX. paup* version 4.0b10 was used to infer the cyanobacterial phylogeny and determine the relative position of the sequenced bands. Modeltest 3.7 was used to determine the optimal base substitution models for the alignment relative to the partial 16S rRNA gene sequences (Posada & Crandall, 1998). Maximum-likelihood analysis was carried out with the parameters derived from the Akaike information criterion (AIC) in Modeltest. The phylogeny was reconstructed by performing a heuristic search with random addition of sequences and a tree-bisection-reconnection branch-swapping algorithm. Bootstrap analysis was performed on the tree topology to evaluate the robustness of the sequence arrangements. Due to computation constraints, bootstrap values were derived under the distance criterion using the ‘fast’ stepwise addition with 10 000 replicates. The selected out-group was the Gram-positive bacterium Clostridium lavalense (strain CCRI-9929, GenBank accession number: EF564278) given that Gram-positive bacteria constitute a phylogenetically supported sister clade to cyanobacteria.

Data treatment

Given the spatial and temporal discontinuities inherent to the sampling strategy, normality in the distribution of the lake data was not assumed and nonparametric statistics applied. Significant patterns in the distribution of meteorological conditions, surface temperature, chlorophyll-a concentration and numerical cyanobacterial abundance were investigated using contingency tables and Kruskal–Wallis analyses. Post hoc analysis consisted in Mann–Whitney U-tests carried out on each pairing, to which Bonferroni correction was applied onto the respective P-values to identify the treatments that significantly differed from each other. The sample size was 50 for all the analyses, unless specified otherwise.

The analysis of DGGE migration patterns relative to the cyanobacterial diversity was carried out sequentially. First, diversity estimates of cyanobacterial GG-I and GG-II were determined for each lake using the Shannon–Weaver diversity index, which was calculated based on the number and peak intensity of DGGE bands observed for each lake. Richness was also defined as the number of DGGE bands detected for each lake. Principal correspondence analysis (PCA) was applied to assess the relationships between water temperature, chlorophyll-a, lake size, proportion of river catchment area covered with wetland, forested and agricultural surfaces and richness of cyanobacterial GG-I and GG-II. The PCA was run for a correlation matrix in a Varimax rotation mode. The data sets were transformed by column standardisation prior to analysis. The significance of DGGE-derived cyanobacterial diversity patterns, based on the PCA associations, was investigated using Mann–Whitney U-tests. Finally, the significance of the occurrence of prominent individual DGGE bands was determined against qualitative descriptors using Fisher's exact tests of independence.

The statistical treatments were performed using PASWStatistic 17.0 module in spss for Mac and Microsoft Excel. The analyses were conducted as nondirectional two-tailed tests.


River catchment topographical attributes: ecoregion delineation

A zonation pattern along the longitudinal axis highlighting three ecoregions in the study area was visible following the analysis of the indices used to describe the typology of the river catchments (Fig. , Table ). Zone I corresponded to the midland area where land use is dominated by agricultural activities. Zone II corresponded to the intermediate band with mixed land cover. Zone III, constituted a band of coastal river catchments, contained well-drained and mountainous terrains with lower anthropogenic pressure.

View this table:

Average index values for the river catchments sampled during summer 2009 within the three ecoregions delineated in west and north-west Ireland

EcoregionAverage index value
SurfaceDrainageAgricultural coverForestry coverWetland coverAltitudeRiver quality

Meteorological conditions, chlorophyll-a distribution and trophic status

Contingency analysis revealed a significant relationship between the meteorological conditions recorded at each sampling site and the individual summer months during which sampling was carried out (X= 10.68, P< 0.05, d.f. = 4). This pattern was supported by the analysis of the meteorological variables recorded at two fixed weather stations within the study area and operated by the national meteorological service (Table ). Surface temperatures in the lakes sampled in July were also significantly higher than those recorded in June and August (Kruskal–Wallis analysis, P< 0.05, H = 11.37).

View this table:

Average values of global radiation, rainfall and wind speed recorded within the study area at the weather stations operated by Met Éireann for June, July and August 2009

Time periodBelmullet weather stationClaremorris weather station
Global radiation (j cm−2)Rainfall (mm)Wind speed (knots)
June (av ± SD)2205 (706)a1.6 (2.9)a7.1 (2.3)a
July (av + SD)1769 (545)b4.4 (6.8)ab7.3 (2.6)ab
August (av + SD)1236 (454)c5.6 (5.1)b8.7 (2.2)b
  • The letters a, b and c show the homogeneous subsets from one-way anova analyses and subsequent Tukey's HSE post hoc tests (P< 0.05).

  • av, average.

At the time of sampling, surface chlorophyll-a concentrations ranged between 1 and 29.2 μg L−1 in the 50 lakes. The low chlorophyll-a concentrations observed in the lakes sampled in June coincided with high water clarity, whereas those sampled in August showed both low phytoplankton biomass and weak light penetration. Even though chlorophyll-a levels were greater in the lakes sampled in July, this was not statistically supported. Considering all the data, chlorophyll-a concentrations were significantly greater in zone I than in zone III (Kruskal–Wallis analysis, P< 0.10, H = 5.77).

The trophic state of the lakes for which this information was not available from EPA Ireland was estimated based on chlorophyll-a and Secchi depth values. Of the 50 lakes sampled, 28% were classified as ultraoligotrophic; 38%, oligotrophic; and 34%, meso-eutrophic, the latter being mostly located within Zone I.

Distribution of cyanobacteria: light microscopy overview

A microscopy-based overview of the abundance of cyanobacteria was carried out based on the enumeration of broad morphological types in the samples collected. The numerical abundance of cyanobacteria was significantly lower in August than in June and July (Kruskal–Wallis analysis, P< 0.05, H = 8.20), corroborating the trend revealed by the chlorophyll-a analysis. However, no significant relationship was observed between cyanobacterial abundance and meteorological conditions or their distribution in the three ecoregions (Kruskal–Wallis analysis, P> 0.10, H < 2.53). Morphotypes of filamentous Oscillatoriales and Nostocales were found in 65% of the lakes with concentrations > 4 × 103 trichomes per litre being observed in eight lakes.

DGGE analysis of cyanobacterial diversity

There was great variation in amplicon migration profiles, both in pattern complexity and band intensity (Fig. ). Hierarchical clustering segregated lakes into two clusters, which were different for cyanobacterial GG-I and GG-II (Fig. ). The number of bands per sample provided an estimate of species richness, which varied from 1 to 23 and 4 to 24 bands for GG-I and GG-II, respectively. The Shannon–Weaver diversity index was also determined based on the relative intensity of the DGGE bands for each sample and was on average significantly lesser for GG-I (median 1.48) than for GG-II (median 1.85) (Mann–Whitney U-test, P< 0.01). Richness and Shannon–Weaver index were positively correlated for both cyanobacterial types (Spearman, P< 0.001, r > 0.72). There were also positive correlations between richness and cumulated band intensity per sample for both GG-I and GG-II (Spearman, P< 0.001, r > 0.63). Overall, no significant pattern was found in the diversity of GG-I and GG-II in the three ecoregions and with the meteorological conditions at the time of sampling. However, combined cyanobacterial diversity was significantly lower in August compared with June and July (Kruskal–Wallis analysis, P< 0.06, H = 5.65), corroborating the numerical abundance analysis.

DGGE migration patterns of partial 16S rRNA gene amplicons for both cyanobacterial GG-I and GG-II (a and b, respectively) in a selection of Irish lakes sampled during summer 2009. The black arrows and numbers under the lanes indicate the bands that were excised and sequenced.

Hierarchical clustering of the lakes sampled during summer 2009 according to their similarity in cyanobacterial GG-I and GG-II (a and b, respectively) community structure. The segregation of the clusters of lakes is also visible in terms of chlorophyll-a concentration.

Mesoscale distribution of cyanobacteria: multivariate analysis

The PCA returned three components that accounted for 68% of the total variance. Component 1 (CP1) was positively related to the proportion of river catchment covered by agricultural surfaces and lake size, while proportions of river catchment covered by forestry and wetland surfaces scored negatively. Cyanobacterial GG-II, chlorophyll-a and water temperature scored the highest on component 2 (CP2), while proportion of river catchment covered by forestry was negative. Cyanobacterial GG-I and to a lesser extent chlorophyll-a contributed positively to component 3 (CP3). The projected score plots of the individual lakes along the components revealed some clustering patterns (Fig. ). In particular, a habitat gradient reflecting the geographical delineation of zones I, II and III was defined along CP1. In the plane defined by CP2 and CP3, interpreted as GG-II and GG-I, respectively, partial segregation of the lakes was visible with respect to chlorophyll-a, water temperature, forest cover of river catchment and lake size. These trends were confirmed with Mann–Whitney U-tests (Table ). In particular, significantly higher diversity of both GG-I and GG-II was found in lakes with chlorophyll-a levels > 2.5 μg L−1. Furthermore, higher diversity of GG-II cyanobacteria was observed in mesotrophic and eutrophic lakes (Mann–Whitney U-test, P= 0.061).

View this table:

Significance of Mann–Whitney U-tests carried out to identify patterns in the diversity of cyanobacterial GG-I and GG-II in relation to the environmental variables included in the PCA analysis

Variable (unit)Median valueGenotype groupP-value (two-tailed)PCA
Lake size (ha)100I0.060*PC1 (+0.62)
Forest cover (%)12.5I0.054*PC1 (−0.60)
Temperature (°C)16.5I0.099*PC2 (+0.65)
Chlorophyll-a (μg L−1)2.5I0.034**PC2 (+0.55)
II0.047**PC3 (+0.50)
  • PCA, principal correspondence analysis.

  • + and − indicate the value for the variable along the corresponding principal component axis.

  • * and ** indicate significance at the 90% and 95% levels, respectively.

Sample projections along the principal components highlighting the ecoregions delineated in the study area (as per Fig. ). The selected variables for the analysis included water temperature, chlorophyll-a, lake size, proportion of river catchment area covered with wetland, forested and agricultural surfaces and richness of cyanobacterial GG-I and GG-II.

Partial 16S rRNA gene phylogeny inference

Randomly selected bands were excised from the gels, re-amplified and sequenced. The blast search returned matches to cyanobacteria only, for all the bands excised, whose positions in the 16S rRNA gene phylogeny were ascertained by maximum likelihood (Fig. 2002). In particular, the Nostocales,Stigonematales and Oscillatoriales-I clades, all corresponding to GG-I, constituted a major assemblage. The remainder of the Oscillatoriales,Pleurocapsales and Chroococcales clades were interspersed within another assemblage into various clusters containing mostly GG-II. Bootstrap values weakly supported intraclade relationships between taxa, but groupings at the assemblage level were not well resolved. The bands sequenced were found present in all the cyanobacterial orders but the Pleurocapsales.

Most likely tree inferred from the maximum-likelihood analysis of 16S rRNA gene sequences. The optimal base substitution model derived from the AIC criterion in Modeltest 3.7 was a TVMef + I + G model with the following constraining parameters for base substitution frequencies, proportion of invariable sites and gamma distribution shape parameter, respectively: A–C = 1.0862, A–G = 2.9378, A–T = 1.5517, C–G = 0.5783, C–T = 2.9378, G–T = 1.0000; I = 0.3180; γ = 0.5160. Numbers on the branches indicate branch frequency from 10 000 bootstrap replicates (values < 50% not included). The selected sequenced bands from Irish lake extracts are indicated in bold. The arrows indicate the clusters in which diversity–abundance relationships of sequenced bands were investigated in Fig. .

Community structure and diversity–abundance relationships

Figure shows for both cyanobacterial GG-I and GG-II the segregation of bands into four quadrants of segment lengths set near the half maxima of the occurrence and band intensity distributions. The presence of significant distribution patterns for these associations, based on band intensity, confirmed the trends observed from the multivariate analysis. Notably, clusters B and D (GG-I) were significantly more abundant in lakes contained in river catchments with substantial forest cover and higher chlorophyll-a concentrations, respectively (Mann–Whitney U-test, P< 0.03).

Segregation of the DGGE bands for both cyanobacterial GG-I and GG-II (a and b, respectively) according to their maximum intensity and frequency of occurrence in the lakes sampled during summer 2009.

The occurrence of prominent individual DGGE bands was also considered (Table ). In particular, band II-78, putatively identified after sequencing and blast search as Woronichinia sp., occurred significantly in lakes where water temperature was lower (Fisher's exact test, P= 0.001). Band I-55 identified as Anabaena sp. was significantly more frequent in late summer and mostly occurred in lakes outside zone I (Fisher's exact test, P= 0.05). Band I-78, also identified as Anabaena sp., occurred significantly in lakes with river catchments with higher forest cover (Fisher's exact test, P= 0.01). Noticeably, the intensities of the Anabaena sp. bands sequenced were significantly greater in the eight lakes where substantial levels of Anabaena morphotypes were numerically observed (Mann–Whitney U-test, P= 0.002).

View this table:

Significant relationship (two-tailed Fisher's exact test) between the presence of selected individual bands and environmental descriptors

BandEnvironmental descriptorblast IDMatch similarity (%)Accession number
EcoregionForest coverTemperatureWeather (rain)Time

The relationships between cyanobacterial richness and abundance, the latter estimated via band intensity, were examined for phylogenetically related sequenced bands for GG-I and GG-II. This was carried out for clusters within the Nostocales (n = 9) and Chroococcales (n = 10) because most sequenced bands matched those genotypes. A significant abundance–diversity correlation was observed within GG-I for the Anabaena–Aphanizomenon clade (Spearman, P< 0.001, r = 0.84), while no such relationship was visible for GG-II with the selected Chroococcales bands (Fig. ).

Scatter plots of DGGE-based abundance of cyanobacterial GG-I and GG-II (a and b, respectively) and richness of Nostocales and Chroococcales groups of phylogenetically related sequenced bands for the lakes sampled during summer 2009. (a) Potential positive diversity–abundance relationship for the Nostocales, while no such relationship seems to be supported for the group of sequenced Chroococcales.


The assessment of water quality is often carried out in lakes under severe anthropogenic pressure or considered important for water abstraction or ecological conservation and restoration (Duan et al., 2007; Papastergiadou et al., 2010). At the regional scale, difficulties reside in the selection of representative water bodies scattered across large geographical areas. Application of sampling strategies based on regional frameworks can optimise site selection and limit sampling efforts in neighbouring areas that may share common environmental attributes (Jenerette et al., 2002; Cheruveril et al., 2008). GIS-based sampling strategies have in recent years been applied to the study of planktonic microorganisms to assess water quality across areas encompassing multiple water bodies (Sliva & Williams, 2001; Catherine et al., 2008; Katsiapi et al., 2012). In this study, a simple semi-random stratified design was applied to investigate the regional distribution of planktonic cyanobacteria. Stratification of the river catchment sample population and phased lake sampling ensured that the selected water bodies were distributed across heterogeneous land cover and hydrogeomorphology and also encompassed potential aestival variation. The analysis highlighted in particular the separation of three ecoregions across a near longitudinal axis in the study area, with the more agricultural midland parts containing greater proportions of water bodies of lesser quality with respect to trophic status. This overall zonation probably implies connection with nutrient enrichment from nonpoint sources, which still constitute a major cause of aquatic pollution in regions of Europe and North America (Smith, 2003; Jeppesen et al., 2005). The assessment and prediction of nutrient diffusion fluxes remain difficult due to their intermittent nature, which is strongly related to seasonal agricultural activity or to precipitation regimes (Carpenter et al., 1998).

The diversity of planktonic cyanobacteria was investigated via 16S rRNA gene DGGE analysis across a region of Ireland containing multiple river catchments. Even though nested PCR-DGGE is a powerful molecular fingerprinting method for detecting low-copy-number genes or specific hierarchical taxonomic groups in microbial communities, it has been criticised for quantitative assessments, as the proportionality between initial template amounts and amplicon concentration can be lost through amplification cycling (Muylaert et al., 2002; Zwart et al., 2005). However, previous work has demonstrated that the number of first-round PCR cycles is essential to limit the bias associated with nested PCR-DGGE, recommending < 20 cycles for maintaining the reliability of quantifications (Park & Crowley, 2010).

Both cyanobacterial GG-I and GG-II were detected by DGGE in the lakes sampled in this study. Light microscopy analysis of broad cyanobacterial taxa was also performed to assess the correspondence between DGGE and numerical abundance data. Overall, both approaches gave statistical support to the greater occurrence of cyanobacteria observed in the lakes sampled in June and July. Even though the presence of Chroococcales genera such as Merismopedia sp., Sphongosphaerium sp. or Microcystis sp. was observed by microscopy, no particular focus was placed on spherical morphotypes due to the relatively low occurrence of colonies and the poor structural integrity of preserved samples. Emphasis was on filamentous cyanobacteria, for which substantial trichome concentrations were found in eight lakes mostly sampled in zones I and II during June and July, suggesting that those morphotypes might develop in lakes of lesser trophic state under clement meteorological conditions. Although morphology analysis and DGGE profiling measure different taxonomic units, the microscopy-based trend was corroborated by the molecular analysis with the distribution of sequenced Anabaena sp. bands.

The large number of 16S rRNA gene sequences available in databases facilitates the phylogenetic classification of cyanobacterial taxa in complex environmental samples. In this study, the designations GG-I and GG-II were based on the sequence homology of the two reverse primers with their targets, as initially designed by Nübel et al., (1997). However, GG-I and GG-II may not be representative of true phylotypes as they are found in rDNA-based phylogenies interspersed in multiple clades of paraphyletic origins, highlighting the genetic similarity of morphologically different genotypes. Prominent sequenced bands returned matches to common cyanobacterial genotypes only, with no heteroduplex products, suggesting that amplification of nontarget sequences was limited. Even though short sequences were used in the alignment, the tree typology largely reflected the assemblages usually obtained (Litvaitis, 2002). The analysis highlighted in particular for the lakes sampled the ubiquitous nature of Chroococcales and Nostocales taxa, which corresponded to the majority of the bands sequenced. However, owing to the fact that < 15% of the band population was sequenced from the DGGE gels, there could have been underrepresentation of some cyanobacterial taxa, such as Pleurocapsales.

Hierarchical classification of the DGGE profiles segregated the lakes for cyanobacterial GG-I and GG-II based on band richness and intensity. Even though relatively weakly supported, statistically, lesser diversity in both GG-I and GG-II was recorded in August, concomitantly with meteorological conditions dominated by precipitations and high-wind regimes. Results also showed at the regional level that lakes classified as mesotrophic or eutrophic generally supported greater diversity of GG-II cyanobacteria. These associations were underpinned by additional relationships, as evidenced from the multivariate analysis. In particular, a majority of GG-I cyanobacteria tended to occur in larger water bodies contained in river catchments covered substantially with forestry surface. Cyanobacterial biomass has however been observed to be lesser in water bodies near forested areas, which might limit diffusion fluxes of nutrients and hence phytoplankton growth (Catherine et al., 2008). Nostocales cyanobacteria such as Anabaena sp. can fix dinitrogen; this attribute can provide them with a competitive advantage over co-occurring phytoplankton species and could explain the pattern observed in the distribution of GG-I cyanobacteria in this study.

The CSR model defining the primary ecological strategies identified among terrestrial plants, based on resource stress and habitat disturbance, has to some extent been satisfyingly applied to phytoplankton organisms (Grime, 2001; Reynolds, 2006). Although probably variable for individual species, the S-strategy nature of buoyant Anabaena and Microcystis can be contrasted to the R-strategist Oscillatoria often associated with density gradients (Mur et al., 1999). The presence of BdII-83 (Oscillatoria), which was linked to both meteorological conditions and the time period of sampling, possibly illustrates the well-documented disruption of stratification by wind-driven turbulence. Classification schemes have been developed to assess functional diversity in the water column based on the presence of clusters of specific phytoplankton organisms, which can be used to evaluate seasonal changes in dominance, responses to eutrophication or the effects of nonseasonal physical forcing (Reynolds, 2006). In particular, sequences of group succession can be recognised for lakes of different trophic state. For example, toxigenic Woronichinia sp. or Microcystis sp. are typically found in nutrient-enriched lakes. As statistically supported in this study, greater diversity of GG-II cyanobacteria was observed in mesotrophic and eutrophic lake samples. The classification schemes under development could be further refined based on the detection of particular genotypes, which can be achieved via molecular fingerprinting or specific molecular assays.

Multiple studies have emphasised the importance of determining both community structure and species identity to unravel aspects of ecosystem functioning (Covich et al., 2004; Downing, 2005; Cardinale et al., 2006). For aquatic ecosystems in particular, additional insights are needed to better comprehend the connectivity between microbial community richness and function (Peter et al., 2011). Microbial genotypes in planktonic assemblages are continuously replaced by fitter ecotypes, and it has been suggested that lineages with higher diversity have higher contribution to ecosystem functioning than those with comparable abundance but lower diversity (Giovannoni, 2004; Peter et al., 2011). Even though this study did not specifically set out to research such aspects, a positive diversity–abundance relationship was observed for sequenced bands of the Anabaena species complex. This may highlight the recruitment of new species via immigration or the increase in previously undetected lineage members (Youssef & Elshahed, 2009). In contrast, such pattern was not found for the sequenced bands of Chroococcales.

Conservation and restoration efforts of water bodies can be carried out on a case-by-case basis but the sustainable management of aquatic resources across multiple watersheds must be considered at a regional level. This highlights the need for providing regulatory bodies with baseline statistics of biodiversity metrics for future enforcement of sampling strategies, monitoring guidelines and parameterisation of ecological models. Studying molecular-based diversity–abundance trends using clone libraries can be cumbersome, in particular in the context of environmental monitoring and the generation of data for multiple sites. New-generation sequencing is a very promising tool, but the cost of analysis still hinders its common use for analysing multiple samples. While new technologies and protocols are refined, a polyphasic methodology to the determination of community structure potentially remains the best approach to elucidate ecological patterns of distribution of particular species and genotypes.


DGGE fingerprinting of 16S rRNA gene amplicons enabled to identify regional variation in planktonic cyanobacterial diversity among lakes interspersed within multiple river catchments, highlighting significant relationships with summer months, lake trophic state and land cover typology. In the context of aquatic resource management, the combined use of GIS-based sampling strategy and molecular methods offers promising prospects for assessing microbial community structure at varying scales of space and time. This approach may provide support to the monitoring programmes implemented in response to the environmental policies introduced to improve water quality and manage aquatic resources.


The authors wish to acknowledge Dr. R. Raine (NUI, Galway) for suggesting improvements in the manuscript. Thanks are due to A. Gill, E. Flood, A. Feeney, R. Ham and K. Kilroy for laboratory assistance. This work was supported by the Irish Environmental Protection Agency through the STRIVE programme 2007–2013 (fellowship 2008-FS-EH-3-S5).


  • Editor: Riks Laanbroek


View Abstract