Introduction
Cyanobacteria are deemed ecologically important for their contributions to global nitrogen fixation, and carbon flux (Paul, 1978; Capone et al., 1997) and their global biomass in the order of 1014 g C (Garcia-Pichel et al., 2003) is a relevant component of both terrestrial and marine biomes. Biotechnologically, they possess a great potential to act as cell factories by virtue of their relatively simple structure, minimal nutritional requirements, and an ability to synthesize a wide variety of metabolites. In this study, we focus on the cyanobacterium L. aestuarii BL J, a representative of an ecologically important species in marine intertidal mats, endowed with an extraordinary capacity to produce H2 (Kothari et al., 2012) and thus, of potential biotechnological interest.
In Nature, L. aestuarii forms extensive microbial mats in many marshes and intertidal mud flats (Horodyski and Bloeser, 1977; Mir et al., 1991; Paerl et al., 1991; Lopez-Cortes et al., 2001) Microbial mats are dense laminated benthic communities of micro-organisms (Stal and Caumette, 1993). They present an environment that is extreme in many respects, with repeated cycles of desiccation and wetting, intense exposure to ultraviolet (UV) radiation, and changing regimes of salinity (as cell may be exposed to hypersaline marine waters to very dilute meteoric precipitation). The intertidal mats that are exposed to desiccation are restricted in their anaerobic components (Rothrock and Garcia-Pichel, 2005). Although, as in most microbial communities, H2 is a key metabolite in interspecies metabolic linking, it rarely accumulates to concentrations high enough to be exported in significant amounts. This has been linked to the diverse populations of potential H2 consumers that inhabit these communities (Ebert and Brune, 1997; Schink, 1997). However, certain intertidal microbial mats, where intense net H2 export occurs (Skyring Gw and Smith Gd, 1989; Hoehler et al., 2001), are an exception. In an earlier report we found that, when subjected to the standard H2 production assays in presence of excess reductants, two different patterns were observed. The strains from marine intertidal microbial mats exhibited higher rates, steady state H2 concentrations and a lack of H2 uptake (we called this Pattern 2 H2 production), in comparison to those from fresh water, which exhibited lower rates and steady state H2 concentrations followed by uptake of most of the produced H2 (Pattern 1, as was known from standard strain of Synechocytis sp. 6803) (Kothari et al., 2012). The fresh water strain Anabaena sp. PCC 7120 also conformed to Pattern 1 hydrogen production. Thus, the cyanobacteria inhabiting the microbial mats (Pattern 2 H2 production) must have evolved extraordinarily powerful hydrogenogenic abilities to produce/sustain hydrogen under the unusually high concentrations of H2 prevailing in their micro-environment. Of the Pattern 2 cyanobacteria, L. aestuarii BL J had the highest rates and reached the highest steady state H2 concentrations (Kothari et al., 2012). Additionally, this strain also displayed an inducible, strong natural hydrogenogenic capacity under dark fermentative conditions (Kothari et al., in preparation). Infact, the rate of fermentative hydrogen evolution in the strain BL J was 10 times higher than that reported for the closely related strain Oscillatoria limosa (=Lyngbya aestuarii PCC 8106) (Heyer et al., 1989). Hence it was of interest to study the genome of this strain, with a special emphasis on the H2 producing system and the ecophysiological constraints imposed by the environment of origin.
Materials and methods
Strains and culture conditions
Lyngbya aestuarii strain BL J, a recent isolate from marine intertidal microbial mats in Baja California (Kothari et al., 2012), was grown in IMR medium set at 3% seawater salinity (Eppley et al., 1968), modified to incorporate a commercially available seawater salt mixture (Instant Ocean), instead of natural seawater, and supplemented with 0.5 μM (final concentration) NiSO4. The strain was maintained in axenic form on IMR media 1% agar plates (since it is less susceptible to contamination than liquid media) at room temperatures and also cryopreserved for long-term storage. Since the strain grew faster in liquid media, it was grown in 250 ml Erlenmeyer flasks, with 100 ml media in presence of 100 μmol photon m2s−1 light at room temperature to obtain cyanobacterial biomass for microscopy and DNA extractions.
Confocal microscopy
A small pellet from liquid culture was washed and resuspended in 300 μl of fresh IMR medium. To stain the DNA, 4′, 6-diamidino-2-phenylindole, DAPI (2 μg/ml final concentration) was added. To stain the exopolysaccharide sheath, Fluorescein-labeled lectin (wheat germ agglutinin; 5 μg/ml final concentration) was added. The preparation was incubated for 1 h in dark at room temperature, and the filaments were washed thrice with fresh IMR medium. Cells were then imaged on glass slides under sealed glass coverslips using a Leica SP5 LASER scanning confocal microscope under a 63X oil immersion objective. Excitation wavelength for DAPI was at 405 nm, excitation for Fluorescein-labeled lectin was at 488 nm, and photosynthetic pigments were excited at 561 nm. The corresponding emissions were detected at 445–465, 520–535, and 675–715 nm. The images presented were maximum Z projections with corrected background (to eliminate background noise). All images were acquired at 1024 × 1024 pixel resolution. All images were manipulated using the image J software suite (Schneider et al., 2012). Imaging of hormogonia, which showed very fast gliding motility, required the use of carbonyl cyanide m-chlorophenyl hydrazine (10 μm final concentration, for 15 min, Santa Cruz biotech) as an uncoupler of proton motive force, to render them immotile.
Transmission electron microscopy (TEM)
Unless stated all steps were under room temperature. Samples were primarily fixed in 2.5% glutaraldehyde in IMR medium for 2 h, followed by four washes in seawater medium over a period of approximately 1 h. Samples were then secondarily fixed with 1% osmium tetroxide in IMR medium for 2 h. Osmium tetroxide was removed by washing with several changes of deionized water over a period of approximately 1 h, followed by block-staining of the cells with 2% aqueous uranyl acetate for 1 h. Uranyl acetate was removed by thorough washing in deionized water. Due to poor preliminary results thought to be caused by incomplete dehydration and resin penetration of the cells, the standard TEM preparation procedure was modified to incorporate increased resident-time in dehydrating agent and epoxy resin, as well as additional gradient steps. This involved: 10 (v/v of reagent grade acetone/deionized water), 20, 40, 60, 80, and 100% anhydrous acetone for four consecutive changes, with each step lasting 30 m. A similar modified approach was employed during infiltration with Spurr's epoxy resin (Spurr, 1969): 10 (v/v of resin/anhydrous acetone), 20, 30, 50, 75%, and four consecutive changes of pure resin. Each step was under rotation for 12 h except for the 10% step, which was 3.5 h. Samples were flat-embedded in fresh resin on Teflon-spray coated glass slides and overlaid with a solid Teflon strip, then polymerized for 24 h at 60°C. Small regions of the pellet were selected and excised from the flat resin layer with a razor, then glue-mounted on a blank resin block in the desired orientation for sectioning. Ultra-thin sections (70 nm) were obtained with a Leica Ultracut-R microtome and collected on formvar-coated 1 × 2 mm slotted copper grids. Sections were post-stained for 5 m with 2% uranyl acetate in 50% ethanol solvent followed by 3 m with Sato's lead citrate (Hanaichi et al., 1986). Images were generated on a Philips CM-12 TEM operated at 80 kV and acquired by a Gatan model 791 slow-scan CCD camera.
DNA extraction, quantification and library construction
Genomic DNA preparations were obtained by PCI (phenol; chloroform; isoamyl alcohol) extraction (Countway et al., 2005). DNA was quantified using fluorometry of ethidium bromide-stained 1% agarose electrophoresis gels and sent for commercial MiSeq 250 Illumina sequencing at the University of Maryland School of Medicine, Institute for Genome Sciences, Genome Resource Center. Genomic DNA libraries were constructed for sequencing on the Illumina platform using the NEBNext® DNA Sample Prep Master Mix Set 1 (New England Biolabs, Ipswich, MA) using the protocol provided, and after DNA fragmentation with an ultrasonicator (Covaris E210). The DNA was purified between enzymatic reactions and the size selection of the library was performed with AMPure XT beads (Beckman Coulter Genomics, Danvers, MA).
DNA sequencing, assembly and annotation
Libraries were sequenced using the 250 bp paired-end protocol on an Illumina MiSeq sequencer. Raw data from the sequencer was processed using Illumina's RTA and CASAVA pipeline software, which includes image analysis, base calling, sequence quality scoring, and index demultiplexing. Data was then processed through both in-house pipelines for sequence assessment and quality control and FastQC (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc/). These pipelines report numerous quality metrics and perform a megablast-based contamination screen. By default, the quality control pipeline assesses basecall quality, and truncates reads where the median Phred-like quality score falls below Q20 (implying more than 99% accuracy in base calling). Data was randomly sampled to create datasets at 100, 150, and 200X coverage. The different datasets were assembled with MaSuRCA (Zimin et al., 2013). The genome assembled by MaSuRCA was subjected to the Institute of Genomic Science (Burja et al., 2001) prokaryotic annotation pipeline forms the core of the IGS Annotation Engine. The pipeline includes gene finding, protein searches, and the pFunc evidence hierarchy that produces automated functional annotation. The output of this pipeline was stored in a Chado relational database and accessed by Manatee for annotation visualization and curation (Galens et al., 2011). The genome was also annotated using RAST -Rapid Annotation using Subsystem Technology (Aziz et al., 2008).
Genomic analyses
The genome annotated by Manatee and RAST was also analyzed manually. Homologs of certain key genes of interest were searched as queries of Psi blast of homologs from phylogenetically close protein sequences from NCBI against the entire genome. Given that the genome is not closed, there is a small probability that the genes reported as missing might be present in the unsequenced part of the genome. The Kyoto Encyclopedia of Genes and Genomes (KEGG) was employed by RAST to gain insight into the various metabolic pathway maps.
Since the bidirectional hydrogenase enzyme was central to the powerful H2 production exhibited by the BL J strain, the architecture of the bidirectional hydrogenase (hox) gene cluster, and hydrogenase accessory genes (hyp) was studied in detail. For the strain BL J, Manatee helped in viewing the genomic organization of the hox and associated ORFs. The physical map of the bidirectional hydrogenase gene cluster and associated ORFs was manually re-constructed to scale in the strains Synechocytis sp. 6803 and Anabaena sp. PCC 7120 (exemplary of Pattern 1), Microcoleus (=Coleofasciculus) chthnoplastes PCC 7420 and L. aestuarii BL J (exemplary of Pattern 2), Lyngbya aestuarii PCC 8106 and Lyngbya majuscula CCAP 1446/4 (closely related to the strain BL J; H2 production capacity unknown). Protein Psi Blast searches were employed to reveal if any ORFs associated with the hox cluster in BL J were also present in any of the other three strains.
To characterize the phylogenetic placement of this strain in reference to other strains in the same cyanobacterial subsection, 16S rRNA sequence (1322 bp) based phylogenetic tree was constructed. The sequences from 83 bacterial species were aligned using ClustalW. The alignment was manually curated and GTR (General Time Reversal model) model with GI (Gamma distributed with Invariant sites) was used to construct maximum likelihood trees with 1000 bootstrap replicates using MEGA 5.2.2 (Tamura et al., 2011).
Bidirectional hydrogenase sequence analysis and protein modeling
The amino acid sequences of the bidirectional hydrogenase from representatives of Pattern 1 (fresh water strains: Anabaena sp. PCC 7120 and Synechocystis sp. PCC 6803) and Pattern 2 (marine intertidal strains: M. chthonoplastes PCC 7420 and L. aestuarii strain BL J) H2 producing cyanobacteria were used for this analysis. The protein sequences of the subunits HoxY and HoxH were individually aligned using Muscle. If the type of amino acid, changed significantly between the two Patterns but remained consistent within a Pattern, it was marked as a significant change (Supplementary Information. 1). The hydrogenase moiety (hoxYH) in Synechocytis sp. 6803 (Pattern 1) and L. aestuarii BL J (Pattern 2) were modeled to study the potential structural importance, of these significant amino acid positions, that might have implications for the function of the bidirectional hydrogenase. Homology models were constructed using the NiFe and NiFeSe hydrogenase templates available in the Protein Data Bank, PDB (Bernstein et al., 1977). Multiple templates were chosen from the PDB for hoxH and hoxY based on the relationship to other bacterial hydrogenases. Both subunits of 1H2A (Higuchi et al., 1997), 1E3D (Matias et al., 2001), 1FRV (Volbeda et al., 1995), 1YQW (Volbeda et al., 2005), and 1CC1 (Garcin et al., 1999) were superimposed using STAMP (Russell and Barton, 1992) and an alignment prepared based on the result using MULTISEQ (Roberts et al., 2006). Each cyanobacterial hydrogenase was profile aligned using CLUSTALW (Larkin et al., 2007) without disturbing the structure-based alignment. The PDB structure files were edited to include protein, the proximal, and medial FeS clusters, the NiFe center, CO, and CN ligands and the Fe or Mg ion at the C-terminus. The previous alignments and edited PDBs were used as inputs for MODELLER (Sali and Blundell, 1993; Eswar et al., 2002) producing 50 independent models. The resulting models were ranked by energy and those with the lowest combined energies were considered in detail. All structures were viewed and figures prepared in Visual Molecular Dynamics software, VMD (Humphrey et al., 1996). The amino acid positions changing consistently between the two patterns were highlighted to study their potential functional significance in the 3D hydrogenase model.
Discussion
Our genomic and cellular description of L. aestuarii BL J shows that this strain shares phylogenetic placement, morphological, and life history traits, with certain other environmentally and biotechnologically relevant cyanobacteria. This clade (Supplementary Information. 2) encompasses globally important marine cyanobacteria like Trichodesmium spp. and globally relevant terrestrial forms such as Microcoleus vaginatus, estimated to be the 3rd and 4th most abundant cyanobacteria on the planet, respectively (Garcia-Pichel et al., 2003). The group includes strains in the genus Arthrospira of importance for large scale production of biomass (Hu, 2004) and used commercially as “Spirulina” as a health food additive (Milledge, 2011) and, of course many other strains of Lyngbya in the so-called Halophilic/brackish/freshwater cluster of biotechnological fame because of their rich and diverse set of secondary metabolites (Engene et al., 2011). All of them being filamentous, non-heterocystous, with discoidal, rather large cells that undergo several rounds of division simultaneously by invagination of cross-walls, and that develop necridial cells for the formation of dispersive hormogonia and for filament separation (Figures 1, 2). Unfortunately given their importance, there are currently no members of this clade for which a system of genetic manipulation has been developed, heavily curtailing biotechnological advancement and bringing and added value to genomic investigations of their members. Given this lack, genomics provides an opportunity to identify and transform genes of interest into other model organisms.
The strain L. aestuarii PCC 8106, isolated from a similar intertidal habitat (a microbial mat in the German Wadden Sea island of Mellum), was the closest to BL J both in terms of the 16S rRNA sequence similarity and phylogeny, and many of the other genomic features (the genome size, percent GC, predicted protein encoding genes and total predicted genes; Table 1). However, significant differences exist between these two strains. Perhaps the most conspicuous being that BL J is almost 50% larger than PCC 8106, with reported cell width around 10 μm (Stal and Krumbein, 1981). Other features like the arrangement of the thylakoid membranes in stacks does not occur in PCC 8196 either (Stal and Krumbein, 1981). We also report the occasional presence of loose, disk-shaped cells within the sheath and the in the media, indicating that cell-to-cell linkages in our strain can be weak, in what can potentially be relevant as an additional means of dispersal that will require focused study.
Our genomic predictions found confirmation in a variety of traits that could be independently assessed. For example, electron microscopy (Figure 2) confirmed the predicted presence of glycogen as a major carbohydrate storage molecule in this strain to the exclusion of polyhydroxyalkanoates. It did also confirm the genomic predictions of cyanophycin synthesis (Figure 2A) and the formation of carboxysomes. Light microscopy revealed the presence of scytonemin in the sheaths of BL J (and we could confirm its preferential synthesis under added UV-A radiation; data not shown), supporting the finding of the entire scytonemin operon in the genome. This all lends credence to other yet to be supported predictions.
A reading of L. aestuarii BL J's genome also speaks directly to some of the environmental constraints of this species in its environment of origin. Known to inhabit exposed intertidal surfaces and the topmost layers of the microbial mats, a high-light phenotype can clearly be surmised from the presence of many photoprotective mechanisms, from extra and intracellular sunscreens, to FPD's that regenerate excess electrons by reducing molecular oxygen to water (Goncalves et al., 2011), to orange carotenoid protein, which helps decouple the light-harvesting systems from the reaction centers (Wilson et al., 2006), as well as from the absence of genes coding for light harvesting pigments that can be considered adaptations to low light intensity like phycoerythrin (Kana and Glibert, 1987), phycoerythrocyanin (Prufert-Bebout and Garcia-Pichel, 1994), or chlorophyll d (Swingley et al., 2008). Intertidal habitats are recurrently exposed to cycles of desiccation and rewetting. Not much is known about the genes involved in desiccation resistance in cyanobacteria but recent transcriptomic studies on the terrestrial strain Microcoleus vaginatus indicate than this is a complex response that involves large sets of genes (Rajeev et al., 2013) and which include complex DNA repair responses, up-regulation of reactive oxygen detoxification mechanisms, the production of osmolites and upregulation of orange carotenoid proteins. Many of the genes involved in these adaptations are also present in this strain, but its mere presence cannot necessarily be linked to desiccation stress resistance. This strain has clearly acquired mechanisms to hold on to moisture, however. In fact, it was very difficult to dehydrate the filaments of this strain for the purpose of TEM preparation. Its thick sheath and the predicted presence of glycine betaine, unusual for a non-halophilic strain, in addition to trehalose, might help in providing desiccation resistance. Finally, a condition typical of the mat habit is that diffusion becomes the major transport mechanism for substrates and products of metabolism. This tends to create diffusion limitations to metabolic activities like photosynthesis and respiration (Garcia-Pichel et al., 1994), which gives relevance to the presence of homologs of genes coding for high affinity carbon uptake protein (hat) and carbon concentrating mechanism (ccm) along with abundant carboxysomes (Figure 2). It also promotes the establishment of anaerobiosis at night within the mat due to consumption by respiration. Under these conditions fermentation of internal reserves though a mixed acid pathway is the only energy-generating metabolism available to strain BL J. Interestingly, this strain lacks the capacity to perform anoxygenic photosynthesis using hydrogen sulfide as an electron donor (homologs of the gene coding for sulfide quinone oxidoreductase were missing) common in microbial mat cyanobacteria. Perhaps this is linked to the low concentrations of sulfide in the upper layers of these intertidal mats compared to mats that are constantly submerged, and where Lyngbya never dominates (Rothrock and Garcia-Pichel, 2005). The presence of recurrent anaerobic conditions will also make soluble ferrous iron available, perhaps leading to the fact that adaptations to iron deficiency such as the products of “iron-stress-induced” gene, isiA; (Straus, 1994; Park et al., 1999) were not detected in BL J.
On the biotechnological potential of this strain we have to note its apparently very rich set of secondary metabolites that range from toxins like Curacin A, Barbamide, Hemolysin-type calcium-binding toxin, to suncreens like scytonemin and mycosporines, to biofuel prospects like heptadecane and pentadecane alkanes. But clearly, biohydrogen is the most promising product of biotechnological importance from this strain (Kothari et al., 2012). Since the standard assays for hydrogen production were performed in the presence of nitrate, a condition in which nitrogenase is not known to be inactive (Ferreira et al., 2009), we rule out the role of nitrogenase in the production of hydrogen. The uptake hydrogenases are known to produce little hydrogen in presence of reduced methyl viologen (Houchins and Burris, 1981). In contrast, reduced methyl viologen is commonly used to assay the bidirectional hydrogenase activity and is likely the enzyme majorly contributing to the strong hydrogen producing capacity of the strain BL J described previously (Kothari et al., 2012). Therefore, the bidirectional hydrogenase gene cluster in this strain is studied in detail with comparisons drawn to other hydrogen producing strains. The organization of the bidirectional hydrogenase (hox) and accessory hydrogenase (hyp) gene cluster was unique in all the four strains (Synechocystis sp. PCC 6803, Anabaena sp. PCC 7120, L. aestuarii BL J, and M. chthonoplastes PCC 7420). A priori, the comparative analysis of the organization of the bidirectional hydrogenase and accessory genes locus revealed no major changes consistent within a Pattern but varying between the two Patterns. A comparative analysis of the organization of the bidirectional hydrogenase locus in the strain BL J revealed that it was similar to that of L. aestuarii PCC 8106 and showed only minor differences with that of L. majuscula CCAP 1446/4 strain (Figure 3).
Interestingly, in all of the Lyngbya strains (Figure 3), a homolog of the hcp gene, predicted to code for hydroxylamine reductase (Wolfe et al., 2002) and typically associated with detoxification of by-products of nitrate reduction (Cabello et al., 2004), is found between genes hoxF and hoxU. Interestingly, in cyanobacteria that are strong hydrogen producers and display sustained concentrations of hydrogen for more than 24 h in dark anaerobic conditions (Ananyev et al., 2008; Kothari et al., in preparation), the hcp gene is present, while it is absent from the genomes of Pattern 1 strains, namely, Synechocystis sp. PCC 6803 and Anabaena sp. PCC 7120. This coincidence may provide a hypothesis worth elucidating the high hydrogenogenic capacity of the Pattern 2 strains. Although typically hcp is annotated as hydroxylamine reductases, it also presents significant homology to known carbon monoxide dehydrogenase (CODH). Perhaps the product of hcp plays a role in the generation of CO needed (Pierik et al., 1999) for the maturation of the NiFe hydrogenases. The source of the CO ligand in the NiFe hydrogenases continues to be unknown (Bürstel et al., 2011).
At the level of hoxYH sequence comparison, we could detect some amino acids substitutions that were consistent within a Pattern but differed amongst the two Patterns. However, none of these amino acids mapped close to the enzyme's active sites, when located on 3D structural models of the hydrogenases of L. aestuarii BL J or Synechocystis PCC 6803 (Figure 4), implying that they are unlikely to modify reaction rates. This suggests that polypeptide differences of the hydrogenase enzyme between the two Patterns are unlikely to explain the functional differences detected previously, necessitating, further study of the biochemistry and regulation of the bidirectional hydrogenase enzymes in these strains. Heterologous expression of the bidirectional hydrogenase from L. aestuarii BL J in model strains such as Synechocystis PCC 6803 might help in gaining a better understanding of the enzyme system.