Samples analyzed in this study (Table 1) were collected during several field trips carried out by the authors in recent years or generously provided by collaborators. Samples collected abroad were fixed in ethanol in situ. Samples collected in France were either fixed or transported directly to the laboratory for immediate DNA purification. Prior to DNA extraction, subsamples of ca. 200 μl volume were taken and, if applicable, ethanol was removed after a short centrifugation step (5 min, 10,000 rpm). Samples were then rehydrated in the initial resuspension buffer of the PowerBiofilm DNA isolation kit (MoBio, Carlsbad, CA USA) and the DNA purified using this kit. DNA was eluted in 100 μl of Tris-HCl, pH 8, and conserved at −20°C. Small subunit ribosomal RNA genes were amplified by polymerase chain reaction (PCR) using primers specifically designed to target the Gloeomargarita and the Thermosynechococcus/Synechococcus clades, including some very basal sequences (colored areas in Figure 1). For the former, we designed the primers 69F-Gloeo (5′-AAGTCGAACGGGGKWGCAA) and 1227R-Gloeo (5′-GATCTGAACTGAGACCAAC) and for the latter, primers 209F-synG9 (5′-TGAGGATGAGCTCGCGGTG) and 1231R-synG9 (5′-GAACTGAGCCRYGGTTTAA). PCR reactions were carried out in 25 μl of reaction buffer, containing 1.5 μl of the eluted DNA, 1.5 mM MgCl₂, dNTPs (10 nmol each), 20 pmol of each primer, and 0.2U Taq platinum DNA polymerase (Invitrogen). PCR reactions were performed under the following conditions: 35 cycles (denaturation at 94°C for 15 s, annealing at 55°C for 30 s, extension at 72°C for 2 min) preceded by 2 min denaturation at 94°C, and followed by 7 min extension at 72°C. Negative (no DNA) and positive (DNA from available targeted cyanobacteria) controls for PCR reactions were used in all cases. DNA from G. lithophora and from S. calcipolaris, T. elongatus, and Synechococchus sp. strains PCC6716 and 6717 was used, respectively, as positive control for the Gloeomargarita and the Thermosynechococcus/S. calcipolaris clades. In cases where no amplification product was obtained, nested or semi-nested amplifications from amplicons obtained with general cyanobacterial primers CYA106F and CYA-1380R were additionally tested. In general, nested or semi-nested amplification attempts failed, suggesting the absence of the targeted sequences from the corresponding samples. 16S rRNA gene libraries were constructed for all positive amplifications using the TopoTA cloning kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. A total of 35 libraries were constructed for Themosynechococcus/S. calcipolaris-like amplicons from different samples and 30 libraries for the Gloeomargarita clade. Clone inserts were PCR-amplified using flanking vector primers, and inserts of expected size were partially sequenced (Beckman Coulter Genomics, Takeley, UK) with reverse primer 1227R or 1231R, yielding sequences of ca. 800 bp. A total of 201 cyanobacterial clones were sequenced (33 for Themosynechococcus/S. calcipolaris and 168 for Gloeomargarita clades). All the sequences obtained were highly similar (>97% identity) but several clones that appeared slightly different were fully sequenced by using forward primers. Complete sequences were assembled using Code Aligner (CodonCode Corporation; www.codoncode.com) prior to phylogenetic analyses. Sequences were deposited in GenBank with accession numbers KJ636536-KJ636761.

Environmental 16S rRNA gene sequences retrieved from our samples were compared with sequences in the GenBank database (http://www.ncbi.nlm.nih.gov/) by BLAST (Altschul et al., 1997). We retrieved the closest sequences found in the database and included them in an alignment containing also sequences from the closest cultivated members and some representative sequences of major cyanobacterial taxa. Sequences were aligned using MUSCLE (Edgar, 2004). Ambiguously aligned positions and gaps were eliminated using Gblocks (Castresana, 2000). The resulting sequence alignments were used as input to build phylogenetic trees by approximate maximum likelihood using Fasttree (Price et al., 2010) with a General Time Reversible (GTR) model of sequence evolution, and taking among-site rate variation into account by using a four-category discrete approximation of a Γ distribution. ML bootstrap proportions were inferred using 1000 replicates. Trees were visualized with FigTree (http://tree.bio.ed.ac.uk/software/figtree/).

Samples of freshly collected aquaria biofilms and microbial mat samples were briefly rinsed in abundant distilled water to avoid the formation of extracellular precipitates upon drying, as previously described (Couradeau et al., 2012, 2013). Then, small sample fragments were deposited onto on formvar-coated transmission electron microscopy (TEM) cupper microscopy grids and let dry. Scanning electron microscopy (SEM) analyses were performed using a Zeiss ultra 55 SEM equipped with a field emission gun. Images were collected in backscattered electron (BSE) mode with a Zeiss Ultra 55 FEG-SEM operating at 10 kV with a 30 μm aperture and a working distance of 7.5 mm using the Angle selective Backscattered (AsB) detector. This mode of imaging provides a contrast which is sensitive to the average atomic number, hence allowing detecting intracellular carbonate inclusions. Elemental compositions of the observed mineral precipitates were directly determined by energy dispersive x-ray spectrometry (EDXS) using an EDS QUANTAX detector and the software ESPRIT (Bruker Corporation, Germany) as previously achieved by Couradeau et al. (2013).

The first step to detect the environmental presence of the two groups of cyanobacteria to which Candidatus G. lithophora and S. calcipolaris G9 belong involved the design specific primers to amplify their respective 16S rRNA genes. These primers targeted a relatively broad diversity of cultured and uncultured cyanobacteria forming a monophyletic cluster with each of them (colored areas in Figure 1), and were validated with control DNA from available cultured species (see Materials and Methods). In the case of Gloeomargarita, our primers were designed to capture the diversity within a large cluster including the only cultivated member of the group, Candidatus G. lithophora, as well as several close environmental sequences, mainly retrieved from thermophilic mats or hot springs, but also from more distant cyanobacteria, including the early-branching operational taxonomic unit (OTU) AQ1-1-1B-84 from Lake Alchichica microbialites (Couradeau et al., 2011) and a sequence from a filamentous thermophilic cyanobacterium (tBTRCCn 301) retrieved from a thermal spring in Jordan (Figure 1). In the case of Ca. S. calcipolaris G9, the clade seemed less diverse, with relatively few related sequences in GenBank. Indeed, the closest sequences to Ca. S. calcipolaris G9 corresponded to those of cultured species, including four highly related Synechococcus species (S. lividus and Synechococcus sp. PCC6716, PCC6717, and OH5) and Thermosynechococcus elongatus, all of which are thermophilic (Dyer and Gafford, 1961; Miller and Castenholz, 2000; Nakamura et al., 2002) (Figure 1). Therefore, in the following, we will refer to this clade indistinctly as the Thermosynechococcus clade or the Synechococcus calcipolaris-like clade.

Because both G. lithophora and S. calcipolaris G9 were enriched from carbonate microbialites of the alkaline Lake Alchichica and because their closest relatives in GenBank had been most often identified in thermophilic microbial mats, we searched to identify these clades in a broad collection of samples collected worldwide that included thermophilic settings compatible with photosynthesis (40–70°C) and carbonate-rich environments, such as microbialites and karstic areas (Table 1). In particular, we explored the presence of the two clades in different microbialite samples from Lake Alchichica collected at different locations along the shore of the lake, time, and depths (from 0 to 15 m depth). We also analyzed microbialite samples from neighboring lakes, including Quechulac, La Preciosa, and Atexcac. We tried to amplify the 16S rRNA genes from the two clades in a total of 62 samples (Table 1). From these, amplicons of the expected size were obtained in a total of 20 samples for the Gloeomargarita clade, whereas only 15 were obtained for the S. calcipolaris-like clade. All of them were cloned and, in the vast majority of cases, the corresponding clone sequences belonged to the targeted groups (see below), showing a high specificity of the designed primers. Only in a few instances, always after nested or semi-nested PCR, were other cyanobacterial sequences retrieved.

The Gloeomargarita group was detected in microbialites from Alchichica, Quechulac, La Preciosa, and Atexcac lakes (Table 1). However, Gloeomargarita identification consistently failed in deep microbialites, being undetected in Alchichica samples from 5, 10, and 15 m as well as in microbialites collected at 15 m depth and kept afterwards in laboratory aquaria. By contrast, Gloeomargarita 16S rRNA genes were amplified from all Alchichica microbialites collected down to 1 m depth, including those maintained in aquaria. They were also amplified from Atexcac, La Preciosa, and Quechulac, but not from all the samples tested (Table 1). Heterogeneous amplification from different samples coming from the same lake was a much stronger trend for the S. calcipolaris-like group. Indeed, its detection in microbialites from Alchichica and surrounding lakes was patchy. This may simply reflect the highly heterogeneous nature of the microbialite environment, which may result in local differences in the distribution of different microbial taxa. However, given that this should also affect the amplification of Gloeomargarita 16S rDNAs, this patchier detection suggests that, although similarly largely distributed in these lacustrine microbialites, the Synechococcus-like G9 clade is less abundant as compared to Gloeomargarita.

In addition to the microbialites in the Alchichica area, the Gloeomargarita clade was also detected in microbial mat samples collected from karstic environments in the Southwest of France (Parc Naturel Régional des Grands Causses and surrounding areas) as well as in a hypersaline microbial mat of the Salada de Chiprana, Spain (Table 1). Interestingly, carbonate precipitation occurs within Chiprana hypersaline mats (Vidondo et al., 1993), most likely due to a combination of photosynthetic and sulfate-reducing activities (Ludwig et al., 2005). Similarly, the S. calcipolaris-like group was detected in mats growing on carbonate rocks from karstic areas in Southern France but also in a microbial mat from El Tatio, a vast hydrothermal field located at high altitude (4300 m) in the Chilean Andes (Table 1). Notably, neither group was detected in plankton samples from karstic systems (encompassing a Mexican cenote and a sample from Southern France) or from the French crater lake Pavin.

Although more prevalent in crater lake microbialites, the identification of the Gloeomargarita and the S. calcipolaris-like groups in samples coming from other, very distant locations, indicates that the distribution of these lineages is cosmopolitan. Ecologically, the two cyanobacterial groups seem to thrive preferentially in thermophilic mats (Dyer and Gafford, 1961; Miller and Castenholz, 2000; Nakamura et al., 2002; Couradeau et al., 2012) or in carbonate rocks, either lake microbialites or carbonates from karstic areas. Interestingly, other deep-branching cyanobacterial groups, such as the Gloeobacter lineage, are also associated with rock environments (Mares et al., 2013).

Surprisingly, even if our specific primers were designed to capture a wide phylogenetic diversity, and even if the sequences obtained came from very different locations, all the 16S rRNA gene sequences that we obtained were all very similar sharing, within each clade, more than 97% identity, the cut-off usually established to define prokaryotic OTUs (Figure 1 and Figure S1). Despite this high similarity, there were some differences possibly reflecting strain variation. Some of this strain variation might reflect local adaptations influenced by geographic and/or physico-chemical differences, both types of factors being very difficult to disentangle. This might perhaps be the case of S. calcipolaris G9-like sequences from Atexcac or Meyrueis/El Tatio or of some Gloeomargarita-like sequences from Chiprana, Alchichica or Quechulac (Figure S1). To really check for strain differentiation, studies using additional, more variable, phylogenetic markers (e.g., multi-locus sequence typing) would be needed.

In summary, the diversity of Gloeomargarita and the S. calcipolaris G9-like group was extremely limited within samples and also across continents and ecosystems. Their low intra-group diversity and the fact that these cyanobacteria (especially the Thermosynechococcus group) did not appear dominant in many of the analyzed samples suggest that they are specialists. These two cyanobacterial lineages provide good examples of the tenet “everything is everywhere, but the environment selects” (Baas-Becking, 1934; Martiny et al., 2006), since they encompass a few cosmopolitan OTUs subjected to strong environmental selection.

Among prokaryotes, it is well known that similar 16S rRNA sequences do not necessarily imply similar phenotypic properties. Therefore, even if we detected 16S rRNA gene sequences very closely related to the two Alchichica intracellularly calcifying cyanobacteria G. lithophora and S. calcipolaris G9, their capability to form intracellular carbonates remained to be shown. Therefore, we carried out scanning electron microscope (SEM) observations on fresh environmental samples where they might be relatively abundant to look for the presence of intracellularly calcifying cyanobacterial cells.

The Gloeomargarita clade is more diverse, but so far only G. lithophora has been shown to possess intracellular carbonate precipitates both in enrichment cultures and in the biofilm it was originally enriched from Couradeau et al. (2012). Sequences affiliating to G. lithophora were also present in Alchichica microbialites, but their abundance in such a complex and diversity-rich environment seemed too low for SEM detection to be feasible (Couradeau et al., 2011). By contrast, confirming original observations, our analysis of wall-biofilm samples from two different aquaria (AQ1 and AQ2) showed a high density of G. lithophora cells containing a huge amount of intracellular Ca, Mg, Sr, and Ba-enriched carbonate precipitates (Figures 2A–C), as confirmed by energy dispersive x-ray spectrometry (EDX) analyses (Couradeau et al., 2012, and data not shown). This remarkable abundance suggests that G. lithophora is particularly fit to this type of biofilm environment and/or that is among the first biofilm-forming organisms on solid substrates (the glass wall in this case). G. lithophora not only forms carbonate inclusions in this semi-natural environment, but seems to have a much higher precipitate content than cultures in BG11 medium (Benzerara et al., in press).

In addition, a recent study of cyanobacterial diversity showed a remarkable relative abundance of sequences (up to 12%) belonging to the Gloeomargarita clade in a calcifying microbial mat from a hot spring at Meskoutine, Algeria (site St7, 55–70°C) (Amarouche-Yala et al., under review). We therefore looked for the presence of cells containing intracellular carbonates using SEM. As can be seen in Figures 2D–H, we identified a number of cells compatible in size and shape with Gloeomargarita sp. and containing intracellular biominerals. EDX analyses on these precipitates showed that they contained carbon, calcium and barium, and lacked phosphorous or sulfur (Figure 2I). This suggests that these inclusions are carbonates of calcium and barium. Barium occurrence in the relatively divergent St7 Gloeomargarita-like cells together with previous observations on G. lithophora (Couradeau et al., 2012) strongly argue for the general capacity of this clade to concentrate barium intracellularly. Our observations suggest that members of the apical Gloeomargarita clade, encompassing G. lithophora and the St7 phylotypes (Figure 1) do form intracellular carbonates in their natural environment. However, whether more basal OTUs in the clade can produce intracellular carbonate inclusions or not remains to be determined.

In the case of the Thermosynechococcus/S. calcipolaris G9 group, all the cultured species examined, including the more basal T. elongatus, do form intracellular calcium carbonates in culture (Benzerara et al., in press). However, we have never been able to observe individual cells with their characteristic polar carbonates in environmental samples. Given the universal ability of members of this clade to precipitate carbonate inclusions intracellularly in culture media, and given that Gloeomargarita cells form large amounts of carbonate inclusions in the natural environments where they are found, it is likely that S. calcipolaris G9-like cells do also produce intracellular precipitates in natural environments. The failure to identify them in natural samples may be related to their low relative abundance in the environments examined. Whether members of this cosmopolitan group are more abundant in other, yet-to-explore ecosystems remains to be determined.

In summary, our study shows that two relatively basal lineages of cyanobacteria producing intracellular carbonates by two different mechanisms are cosmopolitan and exhibit preference for moderately thermophilic microbial mats and microbialites or other, karstic carbonate-rocks. Some of these cyanobacteria, at least members of the Gloeomargarita clade, can be relatively abundant locally. Therefore, their contribution to carbonate precipitation in those ecosystems on the long run may be far from negligible. Furthermore, since these lineages are relatively early-branching (Criscuolo and Gribaldo, 2011; Couradeau et al., 2012; Shih et al., 2013), if their ability to produce intracellular carbonates was ancestral, they may have been important contributors to carbonate production in the geological record (Riding, 2000), especially because thermophilic mats and microbialites were likely more prevalent in the past. The idea that intracellular calcification may have been more important in the past is further reinforced by the recent discovery that this phenomenon is much more widespread than previously thought among cyanobacteria (Benzerara et al., in press). Altogether, these observations lend credit to the hypothesis that extracellular fossilization in cyanobacteria appeared relatively late, perhaps linked to the development of carbon concentrating mechanisms by cyanobacteria, major environmental changes at the surface of the Earth and/or the appearance of conspicuous EPS layers, and might help to explain why cyanobacterial microfossils are not detectable in very old fossil stromatolites (the so-called “Precambrian enigma”) (Couradeau et al., 2012; Riding, 2012).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.