Introduction

The water bloom-forming cyanobacterium Microcystis aeruginosa is a unicellular, colony-forming cyanobacterium distributed worldwide in eutrophic freshwater environments (lakes, ponds, and reservoirs) [1]. The occurrence of Microcystis water blooms has resulted in severe environmental problems including the release of bad odors and bottom layer anoxia. However, the most serious problem associated with Microcystis blooms is the production of hepatotoxic cyanotoxins called microcystins. Accidental exposure to microcystin-contaminated water causes acute poisoning in humans and livestock [2], [3].

To genetically characterize Microcystis isolates in detail, we developed a multilocus sequence typing (MLST) scheme using seven housekeeping loci [4]. Our previous MLST analyses indicated that M. aeruginosa is divided into at least seven distinct phylogenetic clusters [5] with partial correspondence to either colony morphology or microcystin production [6]. On the other hand, the discovery of an intraspecific lineage “group G” is notable because virtually all isolates belonging to this lineage were obtained from a location in Lake Kasumigaura, Japan, in July 2005 [5]. This finding motivates us to collect and genetically characterize many more isolates from this locality and nearby areas to assess whether this group is endemic.

It is widely recognized that recombination highly influences bacterial population structure [7]. The impact of recombination on genetic diversity of bacterial populations was first defined in the milestone work of Smith et al. [8], who proposed the “clonal,” “panmictic,” and “epidemic” population structures on the basis of linkage disequilibrium (LD) between multiple marker loci. Since then numerous microbial species have been characterized with regard to the degree of clonality on the basis of multilocus LD as well as by improved methodologies, revealing differing impacts of recombination on the genetic diversity among species [9], [10]. Recombination has also been suggested to be an important factor in diversification of cyanobacterial species [11]–[13], and several studies have addressed the degree of clonality within them. Studies indicated that the Baltic Sea and North Sea populations of the marine cyanobacteria Microcoleus chthonoplastes [14] are panmictic, whereas the Baltic Sea population of Nodularia [15] and the Japanese population of M. aeruginosa [4] are clonal or weakly clonal. However, a clonal population structure does not indicate the absence of recombination. A substantial level of recombination was observed for M. aeruginosa [4]. Furthermore, because all the above studies were based on the erroneous assumption that recombination occurs equally between within-species individuals regardless of the magnitude of genetic relatedness, it is possible that these analyses underestimated the frequency of recombination by regarding multiple freely recombining units as a single unit. The best way to overcome this inherent problem is by assessing recombination for each distinct intraspecific lineage. However, to date, few studies have assessed the lineage-specific recombination within cyanobacterial species [6].

Accumulation of MLST data allows us to examine group G and other within-species clusters of M. aeruginosa with regard to population genetics, particularly in relation to the impact of lineage-specific recombination and possible endemism. For these reasons, we expanded MLST data sets of M. aeruginosa to include 412 isolates representing 237 unique multilocus sequence types (STs).

Results and Discussion

We obtained 82 novel STs in addition to the 155 previously characterized ones [4]–[6] (see Table S1 for a more detailed description of novel isolates). Phylogenetic analysis of a collection of 237 STs of M. aeruginosa identified seven distinct lineages (groups A–G) with moderate to high statistical support (Fig. 1), which is consistent with our previous analysis [5]. On the other hand, Bayesian assignment analysis (BAPS [16]) identified eight groups; the same seven groups as identified by phylogenetic analysis and an additional group that encompassed all STs belonging to none of the above seven groups (indicated without color coding, Fig. 1). Because possible admixture events using BAPS analysis are more evident in this group than those in other groups (Fig. 1), the ambiguous phylogenetic assemblage of this group is possibly due to recombination between lineages. On the other hand, the result indicated that admixture across lineages appears to be relatively uncommon. This suggests the presence of a barrier to genetic exchange between lineages, although the boundaries are not completely established, as is the usual case for bacteria [17].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Phylogenetic tree of 237 MLST genotypes (STs) of Microcystis aeruginosa.

Bootstrap statistical support was indicated in the major branches. Seven major groups (groups A–G [5]) and a putative hybrid group (group X, suggested by microcystin gene analyses [6]) were color-coded. Note that the hybrid nature of group X was not supported by the result of BAPS analysis of MLST genes. Arrows indicate the panmictic lineage with the number of loci used for LD analysis. The dotted boxes indicate the subset of that group analyzed for LD. The result of admixture analysis is indicated on the right side of the phylogenetic tree where each bar indicates the origin of gene segments of the color corresponding to the phylogenetic group defined by the MLST phylogeny. In addition to the three STs mentioned in the text, ST18 is highlighted in boldface because it represents NIES-843, for which the entire genome sequence is available [45]. The putative ecotypes identified by Ecotype Simulation analysis are also indicated within the black bar. Note that ST49, 99, and 202, which are paraphyletic in the NJ tree, were identified as members of the same ecotype, thus being connected by the black line.

https://doi.org/10.1371/journal.pone.0017085.g001

The differing impact of genetic exchange between and within each lineage was further highlighted by a range of statistical tests for recombination. Previous multilocus LD analysis with a limited number of strains (76 STs) suggested that the population structure of M. aeruginosa is clonal [4]. Overall clonality was again confirmed by analysis with 237 STs in which the standardized index of association (I_A^S [18]), which ranges from 0 (panmixia) to 1 (absolute LD), indicated a significant positive value (P<0.001, Table 1). However, LD analysis of strain subsets revealed the differing impact of recombination within M. aeruginosa. Five intraspecific lineages, groups C–F and a subgroup within group G (designated group G2), were suggested to be panmictic. Exclusion of glnA, which is highly polymorphic in group G (possessing a larger number of alleles relative to other loci; Table S1), extends the range of the freely recombining unit within group G to include more STs (designated group G1) on the basis of I_A^S. Furthermore, results of maximum likelihood tests for tree congruence [9] and estimated parameter values for relative impact (r/m) and rate (ρ/θ) of recombination and mutation supported the recombinogenic nature of the five lineages (Table S2). The average nucleotide diversity (π) for the five lineages of M. aeruginosa (groups C–G) is two to three times less than that for clonal lineages (groups A and B, Table 1), suggesting that barriers to genetic exchange between lineages within M. aeruginosa could be formed in a DNA sequence homology-dependent manner. However, it should be noted that LD analyses of subgroups within groups A and B indicated significant LD, although the genetic divergence within these subgroups is equivalent to that of freely recombining lineages (e.g., group A1). Given that distantly related individuals are often co-isolated from a single bloom (e.g., [5]), physical isolation between each clonal lineage is likely to be absent. Lineage-specific occurrence of genetic exchange, such as phage-mediated transduction in the presence of different cyanophage susceptibility among lineages [19], is thus likely to be responsible for the heterogeneous impact of recombination between lineages of M. aeruginosa. Because the recombination rate is known to decrease log-linearly with increasing genetic distance between individuals [20], [21], the high frequency of recombination observed in multiple intraspecific lineages of an overall clonal species per se is not surprising. A similar pattern has been obtained for other bacteria [22], [23] and has been referred to elsewhere as “cryptic sex” [24]. Interestingly, among the five panmictic lineages, two lines of evidence suggest that group G represents a locally expanding lineage.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Genetic diversity and recombination.

https://doi.org/10.1371/journal.pone.0017085.t001

First, all group G isolates were recovered from a small geographical area, Lakes Kasumigaura and Kitaura that comprise the continuous water body of the Lake Kasumigaura Basin since 2005 (Fig. 2), despite a substantial number of isolates, including more than 30 isolates from Lake Kasumigaura in 2004 (and earlier), being previously characterized (Fig. 3). Yates corrected chi-squared test demonstrated that groups A–F had been evenly recovered from different time points (χ² = 78.86, d.f. = 35, P = 0.03), confirming that the non-availability of group G prior to 2004 is significant. ST30 is the only exception, having been isolated in Nepal in 1988. Given the large genetic distance between ST30 and other group G strains (Fig. 1), ST30 may represent one of the ancestral genotypes of group G. Another exception is a single isolate belonging to group G (ST227) recovered from the Hachiro Lagoon (Fig. 1). Given that the Lake Kasumigaura Basin and Hachiro Lagoon share the same activity of game fishing, the discovery of a closely related genotype at some distance from Lake Kasumigaura might represent a very recent artificial transfer between these two aquatic environments (e.g., mediated by fish stocking or fishing boats).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Distribution of Microcystis aeruginosa STs in Japan.

ST proportions at each sampling location are displayed as pie charts, the size of each being proportional to the number of STs. Intraspecies groups are indicated by different colors as defined in the phylogenetic tree in Fig. 1. Note that the same ST is included in the number of STs in the pie chart indicating pooled samples when they were isolated from different points of time or place. Furthermore, the same ST is not included when it was isolated from the same bloom to avoid the effect of biased sample collections. Red circles on the map indicate locations from where group G strains were isolated whereas blue circles indicate locations from where group G strains were not obtained. Abbreviations for sampling locations are as follows: Ba, Lake Barato; Hg, Hachiro Lagoon; Sw, Lake Suwa; Kg, reservoirs in Kagawa Prefecture; Ok, dams in Okinawa Prefecture; Ta, Takasaki; Is, Iseki; Ts, Tsuchiura; and Ya, Yasuzuka.

https://doi.org/10.1371/journal.pone.0017085.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Distribution of the relative proportions of different phylogenetic groups.

Strains isolated prior to 1990 were arbitrarily pooled to increase the sample size to include at least 18 STs for each bar representation. Strains isolated in 1996–2000 and 2006 were not included due to the small number of isolates available (eight and three strains, respectively). As in Fig. 2, the same ST is included in the number of STs indicated in bar charts when it had been isolated from different points of time or place and not included when isolated from the same bloom.

https://doi.org/10.1371/journal.pone.0017085.g003

Second, group G appears to be highly prevalent in the recent Lake Kasumigaura Basin bloom, i.e., in the presence of group G, strains belonging to other groups have rarely been recovered from the same bloom of M. aeruginosa. Indeed, the recent occurrence of water blooms in summer in the Lake Kasumigaura Basin has been mostly attributed to group G and not to strains belonging to other lineages (Fig. 2). For example, 21 out of 23 STs isolated from Lake Kasumigaura, Takasaki in the early summer of 2005 belonged to group G. A similar pattern was obtained for other time points in Lakes Kasumigaura and Kitaura. At first glance, the bloom in Lake Kasumigaura, Tsuchiura in 2007 appeared to have contained a substantial number of STs belonging to other groups as well as to group G. Given that the dramatic temporal transition of genotypes is known to occur in the population of M. aeruginosa in this lake [5], we consider that this anomaly is due to the inadequate pooling of isolates at different time points from summer to autumn. As expected, an exclusion of October isolates from this data set demonstrated that 10 out of 14 STs belonged to group G, assuring the predominance of group G in this locality in 2007. What is interesting here is that numerous closely related but distinct genotypes in group G were found in a single bloom in contrast to the previously observed pattern of genetic diversity of Microcystis blooms, where multiple, distinct but distantly related genotypes affiliated into different phylogenetic groups co-existed (Fig. 2, [5], [25]). One might think that the observed dominance of group G in a single bloom may be an artifact due to biased samplings favoring the specific detection of group G. Two lines of evidence suggest that this is not the case. First, most strains belonging to group G indicate a unique colony morphology (Fig. S1). Microscopic observation of the bloom associated with group G in the Lake Kasumigaura Basin indicated that majority of Microcystis showed a “group G-type” colony morphology, whereas it has never been observed for isolates obtained from other localities, at least in Japan, excluding the possibility of biased isolation. Second, we used the MA medium [26], an optimum medium for M. aeruginosa isolation that allowed us to successfully recover more than 90% of isolates, excluding the possible biased culture.

Interestingly, the “population snapshot” of group G illustrated by eBURST (a program that can identify the recent divergence of clones on the basis of allele sharing at most multiple loci) [27] clearly indicated the deficiency of clonal complexes (Fig. S2), probably reflecting some genetic discontinuity among the strains. This suggests that the observed genetic diversity of group G represents a long-term (millions of years of) evolutionary consequence rather than a recent diversification. Thus, the most likely explanation of the emergence of group G is that this group is a cryptic ecotype [28]: an ecologically distinct phylogenetic cluster previously present at a low frequency in the Lake Kasumigaura Basin. In response to recent environmental changes favorable to it, group G strains may have dramatically increased in numbers to become a major constituent of water blooms in summer in the Lake Kasumigaura Basin. Indeed, the result of Ecotype Simulation [29] identified groups G1 and G2 as putative ecotypes (Fig. 1). In such a case, there may be further ecological specialization within group G. Interestingly, Group G strains show distinct colony morphology (Fig. S1) that may be beneficial under favorable environmental conditions. Alternatively, the occurrence of group G blooms might have resulted from their recent introduction into Lake Kasumigaura from previously unexplored sources. However, we consider this less probable because a recent introduction of numerous closely related genotypes to a small area is unrealistic and has never been observed for other lineages of M. aeruginosa or documented for other free-living microbes. Taken together, our results suggest that group G represents an expanding lineage in a small geographical area within the past decade. Although a number of cases of a local expansion of pathogenic bacteria have been documented (e.g., [30]), to our knowledge, this is the first report of local expansion of a within-species phylogenetic cluster for aquatic microbes.

Furthermore, frequent recombination within group G is highly suggestive with regard to bacterial population genetics. The emergence of highly adaptive or invasive bacterial clones is often explained by invoking the “epidemic” population structure in which a small number of successful clones dramatically increase their population size against the few background genotypes [8]. Despite the superficial resemblance of the group G-associated bloom to epidemics, the population structure of this bloom is markedly different from that of epidemics. This bloom contains multiple, closely related group G genotypes between which recombination is frequent and a small number of genotypes belonging to different groups between which recombination is rarely observed.

In this study, we have identified multiple panmictic lineages within M. aeruginosa. Because recombination can function as a cohesive force to limit the divergence of each lineage, the single species M. aeruginosa can be recognized as an assemblage of multiple, independent phylogenetic units whose evolutionary consequence could also be distinct. Among these, group G is notable for both basic evolutionary interests and local environmental concerns because it is predominantly responsible for the recent occurrence of water blooms in a restricted region in Japan, although all group G isolates are fortunately nontoxic (Tanabe et al., unpublished data). Although available data suggests that group G is a possible adaptive ecotype, caution is required because a sequence-based cluster can be formed in the absence of ecological selection where the frequency of recombination is negatively correlated to DNA sequence divergence [31]–[33], which is the case for M. aeruginosa including group G. Furthermore, a possibility that group G was formed as a result of geographic isolation and neutral genetic drift exists. Endemism has long been considered unusual for free-living microbes [34], until possible endemic clades were reported for cyanobacteria [35]–[37]. Because our data are largely limited to Japanese isolates, culture-independent methods specifically to detect group G strains from blooms worldwide would be of great help in dealing with this issue. In any event, ecological characterization is mandatory to confirm the ecotypic status of group G. Long-term examination of genetic diversity of group G and other groups would provide us with a unique opportunity to investigate the evolutionary consequence of this lineage in light of various bacterial speciation theories [38]–[39].

Materials and Methods

Supporting Information

Acknowledgments

We thank Fumie Kasai for her assistance for sample collection and helpful discussions. We also thank Wichien Yongmanichai, Yuwadee Peerapornpisal, and Uk On Norong for helping us collect samples in Thailand and Cambodia.