The original strain was isolated from the sulfidic water column of Solar Lake, Israel (Cohen et al., 1975b), and was kindly provided by A. Oren for culturing. A monoalgal culture was grown in modified Chu’s 11 in Turks Island Salts medium at room temperature (average 22.0°C) and ambient light in a 125 mL Erhlenmeyer flask. We extracted whole community DNA using the MPBio FastDNA SpinKit and Fastprep-24 Bead Beater (MP Biomedicals, Solon, OH, USA) following the default protocol, except that 0.3 g of beads were used for bead beating. DNA was quantified using Quant-IT PicoGreen (Invitrogen, Grand Island, NY, USA) and submitted to the University of Michigan DNA Sequencing Core for library preparation and Illumina HiSeq 2 × 100 bp paired-end sequencing.

Using wrappers provided at https://github.com/Geo-omics/scripts, reads were dereplicated with custom perl scripts, trimmed using Sickle (version 1.33) (Joshi and Fass, 2011), and assembled using IDBA-UD (version 1.1.1) (Peng et al., 2012) with the following parameters: –mink 65 –maxk 85 –step 10 –pre_correction. Tetranucleotide frequency was used to bin scaffolds by emergent self-organizing maps (Dick et al., 2009), with a minimum contig length of 500 bp and a window size of 10,000 bp (Supplementary Figure S1). The cyanobacterial bin of scaffolds >1000 bp in length was submitted to the Integrated Microbial Genomes Expert Review (IMG-ER) automated pipeline from Joint Genomes Institute (JGI) for annotation of genes and pathways (IMG accession number: 2660238729; Supplementary Table S1). Raw reads, assembled scaffolds, and gene annotations were submitted to NCBI (project: PRJNA302164).

Phylogenetic analyses of genes of interest (Table 1) was performed with maximum likelihood and the PROTGAMMAGTR algorithm in RAxML 8.1.15 and bootstrapped 1000 times (Stamatakis, 2014). psbA. sqr, and nifHDK gene phylogenies included 44 cyanobacterial isolate genomes and genomic bins from Hot Lake and Middle Island Sinkhole cyanobacterial mat metagenomes (Cole et al., 2014; Voorhies et al., 2016; Supplementary Table S2). Each of these cyanobacterial genome has at least one sqr gene, and 30 of the 44 genomes have a nifHDK gene set. sqr and nifHDK genes from Sulfuricurvum kujiense YK-1 DSM 16994, and other bacterial sqr genes (Supplementary Table S3) were included for context. Translated genes (amino acid sequences) were aligned with clustal-omega (version 1.2.0) (Sievers et al., 2011). Alignments for nifH. nifD, and nifK were concatenated, and the concatenated alignment was used for analysis. Halothece halophytica chlLNB genes were used to root the nifHDK tree (Boyd et al., 2015). Flavocytochrome c:sulfide dehydrogenase (FCSD) genes formed an outgroup for the sqr tree (Marcia et al., 2010a). The most divergent psbA from Gloeobacter kilaueensis JS-1 was used as an outgroup for the psbA tree (Cardona et al., 2015). Phylogenetic trees were drawn with FigTree version 1.4.2 (Rambaut, 2012).

The 4.77 Mb draft genome of Geitlerinema sp. PCC 9228 (henceforth “Geitlerinema”) contains 3,969 protein coding genes on 195 scaffolds. Coverage is on average 905x across the genome. The genome has 100% of universally conserved bacterial genes expected to be present (Raes et al., 2007; Alneberg et al., 2014).

Geitlerinema has key genes coding for proteins involved in photosynthetic and respiratory electron flow, including photosystem II (psbADBCEFO), succinate dehydrogenase (sdhABC), type-1 NADPH dehydrogenase (ndhA-M), cytochrome b6f (petADBCEJ), photosystem I (psaABCDEFK), and cytochrome c oxidation (coxABC) (Supplementary Table S1; Mulkidjanian et al., 2006). It has genes for the Calvin-Benson-Bassham cycle, carbon dioxide concentrating proteins (ccmK₁K₂MN), and Rubisco large and small subunits (rbcLS) for carbon fixation (Figure 1). Superoxide dismutase genes (sodC and sodN) to cope with superoxide formation during photosynthesis and aerobic respiration are also present in the genome. In ancestral cyanobacteria, these enzymes would have been critical for defense against increased production of reactive oxygen species alongside increasing O₂ fluxes (Blank and Sánchez-Baracaldo, 2010; Fischer et al., 2016).

Geitlerinema has genes encoding a complete tricarboxylic acid (TCA) cycle, including 2-oxoglutarate dehydrogenase and succinyl CoA synthase to link synthesis of 2-oxoglutarate through succinyl-coA with succinate (Steinhauser et al., 2012). The genome also has the genes for acetolactate synthases and succinate-semialdehyde dehydrogenase to interconvert 2-OG and succinate through succinic semialdehyde in an alternative closure to the TCA cycle (Zhang and Bryant, 2011). Though it has shc, encoding for squalene-hopene cyclase, Geitlerinema lacks the hpnP gene for hopanoid methylation (Ricci et al., 2015). 2-methylhopanes, derived from 2-methylhopanoids, have been used as a bacterial biomarker in the geologic record (Summons and Lincoln, 2012). Geitlerinema has genes for acyl-ACP reductase and fatty aldehyde decarbonylase, key enzymes in an alkane biosynthesis pathway unique to cyanobacteria (Schirmer et al., 2010; Coates et al., 2014).

In the chlorophyll synthesis pathway, we identified both the aerobic oxidative ester cyclase chlE common to all cyanobacteria (Mulkidjanian et al., 2006), and the oxygen independent ester cyclase bchE. Functional bchE is common in anoxygenic phototrophic bacteria such as green sulfur bacteria and heliobacteria (Sousa et al., 2013), and is only rarely present in cyanobacteria, such as Synechocystis sp. PCC 6803 (Minamizaki et al., 2008). Geitlerienema has the light-dependent NADPH-protochlorophyllide oxidoreductase that produces chlorophyllide in the penultimate step of the pathway. This gene originated in cyanobacteria and is limited to cyanobacteria and phototrophic eukaryotes (Blankenship, 2002; Yang and Cheng, 2004). However, the genome lacks the light-independent oxidoreductase chlLNB, which allows for chlorophyll synthesis in the dark and is observed in all photosynthetic phyla (Blankenship, 2002). Due to their homology, Geitlerinema’s genes for nifHDK are the closest matches to H. halophytica’s chlLNB. Coverage estimates for nifHDK genes are consistent with the rest of the genome, indicating there was no mis-assembly of chlLNB reads into nifHDK genes.

The cyanobacterial genome has genes for a variety of pathways of nitrogen acquisition, including an ammonia transporter gene amt, cyanate lyase cynS, nitrate/nitrite transporters narK and focA, and nitrate assimilation-related genes nitrate and nitrite reductases nirA. nirC, and narB (Figure 1). The cyanobacterium also has urea transporters (urtABCD), urease genes (ureABCDFG), and genes for transport of neutral, branched, and polar amino acids.

Geitlerinema has a comprehensive operon for nitrogen fixation (nifVXSU, nifHDKEB; Raymond et al., 2004). Additional nitrogenase-related proteins are located on the same contig (iscA. nifI₁I₂, ferredoxin, nifN; Figure 2; Boyd et al., 2015). iscA is commonly observed in aerobic diazotrophs, and its recruitment into the genome is linked to a transition to aerobic lifestyle (Boyd et al., 2015). The cyanobacterium has glnB, a member of the PII signal transduction protein family that regulates nitrogen-related proteins, and ntcA, which controls expression of glnB (Forchhammer, 2004). The nifI₁I₂ gene is also a member of the PII protein family, but is characteristic of diazotrophic anaerobes that regulate their nitrogenase activity post-translation, such as Desulfovibirio and Clostridium (Boyd et al., 2015).

Phylogenetic analysis shows that nifHDK genes from Geitlerinema clusters with those from two cultured cyanobacterial genomes (Pleurocapsa sp. 7327, Microcoleus chthonoplastes) and a cyanobacterial genome-from-meta genomic bin (Phormidium OSCR; bootstrap = 100; Supplementary Figure S2). Like Geitlerinema, their nitrogenase operons also hold nifI₁I₂. Geitlerinema has a bidirectional NiFe hydrogenase gene set (hoxEFUYH and hoxW) with its transcriptional regulator (lexA) and hydrogenase maturation proteins (hypBAEDC) (Figure 1). Unlike Geitlerinema, typical nitrogen-fixing cyanobacteria have an uptake hydrogenase (hupSL) to consume H₂ produced in nitrogen fixation. hupSL is under similar transcriptional regulation as nitrogen acquisition genes like dinitrogenase, making expression of hupSL dependent on nitrogen limitation (Tamagnini et al., 2007). In contrast, the bidirectional hydrogenase can be present in both diazotrophic and non-diazotrophic cyanobacteria and is expressed under more diverse conditions. It may, for instance, be used in fermentation or to direct electrons during photosynthesis (Tamagnini et al., 2007). Geitlerinema produces hydrogen during sulfide-dependent AP in the presence of bioavailable nitrogen, as well as in the absence of CO₂, suggesting the nitrogenase-independent hox gene set may be responsible for hydrogen evolution (Belkin and Padan, 1978; Belkin et al., 1982).

The core proteins in photosystem II are encoded by psbA and psbD (Blankenship, 2002). Geitlerinema has one version of psbD and three versions of the psbA gene (Figure 1). The standard psbA, hereafter referred to as psbA3, is in a large clade of typical oxygenic psbA designated “group 4” after (Cardona et al., 2015) (Figure 3). All but one of the isolate genomes in this analysis have at least one copy of this form of psbA, which is used in OP in aerobic conditions (Supplementary Table S2). Coverage estimates and paired-end information suggest that Geitlerinema has two copies of psbA3. It is the sole gene on its contig, and pairs of reads that map to the ends of the psbA3 match to the ends or beginnings of four other contigs. Those portions have 100% identity to the ∼80 bp beginning and end of psbA3. The De Bruijn graph-based assembly algorithm used in this analysis frequently prematurely assembles identical copies of genes (Nagarajan and Pop, 2013). The approach accurately assembled the two copies of psbA3, but could not automatically bridge the copies.

One additional version of psbA, henceforth referred to as psbA1, has 85% sequence identity to psbA3 (blastx). psbA1 is located in a different operon but on the same scaffold near ntcA and photosynthetic subunits psbN and psbH. psbA1 is in a well-supported clade with genes from Geitlerinema sp. PCC 7105 (2510100750) and Geitlerinema sp. BBD 1991 (BBD_1000995126; bootstrap = 81; Figure 3). This gene is also in a larger “group 2” (Cardona et al., 2015) that includes psbA genes from cyanobacterial genomic bins (MIS_1001011011, MIS_100039089) sourced from a low-oxygen cyanobacterial mat in the Middle Island Sinkhole (Voorhies et al., 2016). Group 2 genes encode D1 proteins used under growth conditions that do not favor water oxidation (Cardona et al., 2015), such as heterotrophy in the dark (Park et al., 2013), photosynthetic electron transport inhibition (Kiss et al., 2012), or oxygen-sensitive processes such as nitrogen fixation (Toepel et al., 2008; Wegener et al., 2015). Finally, the third variant, psbA2, is 59–88% similar to psbA1 (discontinuously; tblastx), and 90% similar to psbA3 (over 99% of their lengths; blastx). psbA2 is in a clade with genes from a Middle Island Sinkhole cyanobacterium (MIS_100299244), Geitlerinema sp. BBD1991 (BBD_1000995127) and Phormidium sp. OSCR (2609132164; bootstrap = 94). Proteins from these “group 3” psbA genes (Cardona et al., 2015) are expressed under microaerobic conditions, with modified electron transfer to cope with the changing redox environment (Figure 1; Sicora et al., 2004, 2009; Sugiura et al., 2012).

Two copies of the gene for sulfide quinone reductase, sqr, are in the genome of Geitlerinema (Supplementary Figure S4). This gene is involved in the oxidation of sulfide for detoxification or to harvest electrons for AP, such as in purple and green sulfur bacteria (Theissen et al., 2003; Marcia et al., 2009, 2010a; Gregersen et al., 2011). Each sqr is located upstream of arsenic resistance arsR-type genes, putatively involved in transcriptional regulation of sqr under conditions such as sulfide exposure (Nagy et al., 2014). One sqr, referred to as sqr1 hereafter, matches the previously cloned and sequenced sqr gene of Geitlerinema (Bronstein et al., 2000). This enzyme mediates the reduction of plastoquinone and oxidation of sulfide in sulfide-dependent AP (Arieli et al., 1994). On the sqr phylogenetic tree, Geitlerinema’s canonical sqr groups with other cyanobacterial versions that are considered type I (Marcia et al., 2010a; Gregersen et al., 2011; bootstrap = 100; Figure 4; Supplementary Figure S5). Hydrogen sulfide affinities for cyanobacterial sqr in this cluster are high, with K_m in the micromolar range (Arieli et al., 1991; Bronstein et al., 2000). Geitlerinema. Coleofasciculus (formerly Microcoleus), Halothece, and proteobacterial members in this cluster have been shown to grow with this sqr on sulfide-induced AP (Oren and Padan, 1978; Cohen et al., 1986; Jørgensen et al., 1986; Schütz et al., 1999).

The second sqr gene, referred to as sqr2 hereafter, is 25–50% identical over discontinuous fragments to the sqr1 gene (tblastx). It is most similar (tblastx 48–63% identity to overlapping fragments that span the length of the gene) to that of Chloroherpeton thalassium ATCC35110, a green sulfur bacterium, with similar identity to the sqr genes of other sulfur oxidizing bacteria. sqr2 clusters phylogenetically with two other cyanobacterial sqr [Geitlerinema sp. PCC 7105 and BBD 1991 (bootstrap = 97)], among a group of known type VI versions (bootstrap = 100; Figure 4; Supplementary Figure S5). They include sqr from thiotrophic proteobacteria (Marcia et al., 2010a) and green sulfur bacteria (Gregersen et al., 2011), including that of Chlorobium tepidum used in sulfide oxidation when sulfide exceeds 4 mM (Chan et al., 2009).

Arsenic resistance genes arsB (an arsenite efflux pump) and arsC (arsenate reductase; Slyemi and Bonnefoy, 2011; van Lis et al., 2013), and arsR, the putative regulatory protein of sqr, are in an arsRBC operon downstream of sqr2. Upstream of sqr2 are genes encoding glutathione S-transferase and a multidrug efflux pump (Supplementary Figure S4). Another arsB is located downstream of sqr1 but in a different operon. In combination with arsB, two arsenite-tranporting ATPases arsA genes are present in the genome, one located on the same scaffold as the nitrogenase gene suite and the other on the longest scaffold in the dataset. Geitlerinema does not have known arsenite oxidizing genes (aioAB. arxA) or respiratory arsenate reductase genes (arrA), based on gene annotation and BLAST searches against known genes (Hoeft et al., 2010; Slyemi and Bonnefoy, 2011). It has an annotated chromate transporter (chrA) on another scaffold that has 52% positive match (blastp) to a similarly annotated gene in Synechocystis sp. PCC6803 that functions as an arsenite uptake transporter for arsenite oxidation (Nagy et al., 2014; Figure 1). On another scaffold, a gene belonging to the DUF302 protein superfamily of unknown function is quite similar (75% positive, 56% identical blastp) to a Synechocystis gene in the sqr1 plasmid operon, as well as to a potential arsenic oxidase gene in Agrobacterium (61% positive, 44% identical, blastp; Nagy et al., 2014). Geitlerinema also has a methyltransferase similar (84% positive, blastx) to the arsM gene of Synechocystis used in arsenite methylation for detoxification (Yin et al., 2011).

Our results indicate that Geitlerinema sp. PCC 9228 and other sulfide-tolerant and/or sulfide-using cyanobacteria have multiple different types of psbA genes that are inferred to meet physiological and biochemical needs under changing light and redox conditions. The psbA gene encodes the D1 protein that directly supports the water-oxidizing cluster in PSII (Ferreira et al., 2004; Fischer et al., 2015). Due to its role in water oxidation, this protein experiences high levels of oxidative damage and degradation. By impacting water oxidation, changes in redox such as sulfide inhibition of the water oxidizing complex in PSII (Miller and Bebout, 2004) and/or non-optimal light conditions lead to excessive energy (Klatt et al., 2015b) and influence the risk and magnitude of oxidative damage to D1 (Aro et al., 1993). Thus, cyanobacteria that experience such dynamic conditions may use another of the multiple psbA genes in their genomic repertoire under different light and oxygen regimes to mitigate oxidative damage (Schaefer and Golden, 1989; Campbell et al., 1999; Sicora et al., 2009; Gan et al., 2014; Cardona et al., 2015; Ho et al., 2016). Geitlerinema has two copies of psbA for OP under high oxygen and/or light levels (both psbA3; group 4), one gene for non-oxygen evolving PSII (psbA1; group 2), and one gene for microaerobic conditions (psbA2; group 3; Figures 1 and 3). We infer that these distinct photosynthetic genes reflect the adaptation of Geitlerinema to a variably lit, low-oxygen, and sulfidic lifestyle.

During high light conditions, when Geitlerinema conducts OP (Cohen et al., 1975a, 1986), the psbA3 (group 4) genes are likely used to reduce photoinhibition and oxidative stress, as observed in Synechococcus sp. PCC 7942, Synechocystis sp. PCC6803, and Thermosynechococcus elongatus (Schaefer and Golden, 1989; Sicora et al., 2006; Kós et al., 2008; Sugiura et al., 2010). Other cyanobacteria synthesize identical D1 proteins from different genes at high light versus regular light conditions (Sugiura et al., 2010). Two copies of psbA3 in its genome may equip Geitlerinema to continue OP in times of sufficient and/or high light, such as when mixed up into the epilimnion or in holomixis. Given that psbA3 clusters with standard and high-light variant genes (Figure 3), it remains to be seen if one or both of Geitlerinema’s two copies of psbA3 is cued to the different light intensities the organism may experience in its natural habitat.

The psbA1 and psbA2 genes likely provide Geitlerinema with the flexibility to conduct photosynthesis and nitrogen fixation under varying oxygen and/or sulfide concentrations. With sufficient light, the cyanobacterium may rely upon psbA2 (group 3), which is keyed for microaerobic conditions. Changing redox environment promotes synthesis of psbA2-like variants (group 3) in Thermosynechococcus elongatus (Sugiura et al., 2012) and Synechocystis sp. PCC6803 (Sicora et al., 2004). However, light levels are lower than required for OP in the hypolimnion of Solar Lake, the prime habitat of Geitlerinema. Together with continuous sulfide exposure, these conditions often favor AP in cyanobacteria (Klatt et al., 2015a, 2016) and in green sulfur bacteria (Hanson et al., 2013; Findlay et al., 2015). Groups 2 and 3 psbA genes in the current study are in close chromosomal proximity to sqr genes (Figure 3), suggesting an intriguing but currently untested linkage between modified PSII and sulfide exposure. Alternative D1 proteins have been hypothesized to be directly involved in the transfer of electrons from donors other than water (Murray, 2012), such as in sulfide oxidation during nitrogen fixation (Becraft et al., 2015; Olsen et al., 2015). However, whether alternative D1 proteins are directly participating in this unusual photochemistry or serve to inactivate PSII in AP, remains to be tested.

Very little sulfide is required to limit OP in Geitlerinema (Cohen et al., 1986), and in the transition between oxygenic to AP, continuous exposure induces synthesis of proteins such as SQR (Arieli et al., 1991). SQR has a higher affinity for plastoquinone than PSII (Klatt et al., 2015a), thus in this induction period and with continual sulfide exposure, Geitlerinema may also reformulate its photosystem II machinery to reduce oxygen production. Finally, the psbA1 (group 2) gene is also likely used to disable oxygen production during times of nitrogen fixation. Non-oxygen producing group 2 enzymes have been implicated as structural ‘placeholders’ that may allow oxygen-sensitive processes such as nitrogen fixation to occur (Sicora et al., 2009; Murray, 2012; Wegener et al., 2015). Further experiments targeting gene and protein expression will confirm the physiological roles of the variant psbA genes.

Variant psbA genes such as those observed in Geitlerinema likely permitted ancestral and potentially AP-capable cyanobacteria to meet metabolic requirements in dynamic physicochemical conditions. The importance of AP cyanobacteria and their metabolisms on Archean and Proterozoic oxygen levels is linked to the timing of OP. Mat-forming cyanobacteria in modern hypersaline and hot spring ecosystems likely face the same environmental challenges as their stromatolite-forming ancestors (Stal, 1995; Grotzinger and Knoll, 1999). As in modern systems, the metabolisms and mat-building lifestyles of ancestral cyanobacterial populations would have promoted light and redox dynamics on spatial (depth) and temporal (diel) scales (Canfield, 2005; Blank and Sánchez-Baracaldo, 2010; Lalonde and Konhauser, 2015; Sumner et al., 2015). Genomic and physiological studies on contemporary mat-forming cyanobacteria and their responses to changing environmental parameters, such as different versions of psbA keyed to light and/or oxygen levels, inform potential genetic and physiological strategies in ancient cyanobacteria. Given the long geologic history of variable and low atmospheric oxygen concentrations, the first appearance of alternative psbA in cyanobacteria is uncertain. The functional differences and potential heterotachy in these homologous genes dictate caution when evaluating their timing and evolutionary order. However, the basal arrangements of group 2 to group 3/4, and group 3 to group 4 in the phylogenetic tree hint at an ancestral development of water oxidation when atmospheric oxygen was low (Cardona et al., 2015). The perpetuation of diverse and evolutionarily old psbA genes in cyanobacterial genomes allow for efficient metabolic functioning regardless of oxygen levels/needs (Murray, 2012). These variant genes were likely retained in cyanobacterial genomes for oxygen-sensitive processes, such as nitrogen fixation, in an increasingly oxidizing environment.

Geitlerinema has a suite of genes for uptake of nitrogen (nitrate, nitrite, urea) as well as nitrogen fixation through a comprehensive nitrogenase gene suite (Figures 1 and 2). Sulfide-dependent AP was measured in select strains grown in nitrogen-replete media, including Geitlerinema sp. PCC 9228, Coleofasciculus chthonoplastes, and Pseudanabaena FS39, suggesting nitrate assimilation occurs under AP conditions (Cohen et al., 1986; Klatt et al., 2015a). Geitlerinema is also capable of AP-dependent nitrogen fixation (Belkin et al., 1982), potentially using sulfide to scavenge residual oxygen or donate electrons to nitrogenase (Stal, 2012).

Clues about the cyanobacterial transition from an anaerobic to aerobic lifestyle are apparent in regulatory genes for nitrogen acquisition. Due to its evolution under anoxia and strict requirement for anaerobic conditions, cyanobacterial diazotrophy in an increasingly oxidizing environment required new adaptations (Boyd et al., 2011; Stüeken et al., 2015). In the nif operon of Geitlerinema is nifI₁I₂, which encodes a signal transduction protein of the PII family that is present in diazotrophic archaea and select anaerobic bacteria, but has not been studied in cyanobacteria (Forchhammer, 2004; Boyd et al., 2015). In those organisms, NifI₁I₂ inhibits nitrogenase activity when the cell is no longer nitrogen limited (Kessler et al., 2001). Geitlerinema’s nifHDK gene suite is phylogenetically grouped with those from three other cyanobacterial genomes that also have nifI₁I₂ in their nitrogenase operons (Supplementary Figure S2), and members of this cluster are adapted to low-oxygen and/or sulfidic conditions. C. chthonoplastes can continue to operate OP at low sulfide concentrations and is also capable of sulfide-dependent AP (Cohen et al., 1986). Similar to Geitlerinema and its psbA1. Pleurocapsa sp. PCC 7327 also has a psbA gene that codes for a rogue D1 protein typically used in anoxic conditions (Wegener et al., 2015).

Sulfide quinone reductase (SQR, encoded by sqr) oxidizes sulfide to elemental sulfur, and this process can be coupled with lithotrophic or phototrophic growth in bacteria and archaea or detoxification of sulfide in eukaryotes (Theissen et al., 2003; Marcia et al., 2010a). The genome of Geitlerinema sp. PCC 9228 holds two sqr operons. The first operon has the well-studied high-affinity sqr1 (type I), which is translated after a few hours of exposure to micromolar levels of H₂S (Oren and Padan, 1978) and enables Geitlerinema to grow by sulfide-based AP (Cohen et al., 1986). The transcriptional regulatory gene arsR, located downstream of sqr1, likely controls its expression (Nagy et al., 2014). Though lateral gene transfer has distributed various sqr types among bacteria, archaea, and eukarya (Theissen et al., 2003), type I sqr genes in cyanobacteria such as Geitlerinema sqr1 form a distinct and well-supported cyanobacterial phylogenetic subclade within a bacterial clade (Figure 4). This cyanobacterial clade includes genes from AP-capable or sulfide-tolerant members such as C. chthonoplastes. Synechocystis sp. PCC6803, H. halophytica, and Geitlerinema sp. BBD (Cohen et al., 1986; Theissen et al., 2003; Marcia et al., 2010a; Nagy et al., 2014; Den Uyl et al., 2016). SQR is an evolutionarily ancient enzyme that was widespread in organisms during the Proterozoic (Theissen et al., 2003), during which the oceans had variable redox conditions including euxinia. Although the evolutionary history of sqr in cyanobacteria appears to be complex and remains unresolved, the sqr1 gene is thought to be endogenous to cyanobacteria (Theissen et al., 2003). As a critical component of sulfide-based AP, this gene would have enabled ancestral cyanobacteria to thrive during the Proterozoic, when periods of photic zone euxinia would have favored AP and tempered oxygen production (Johnston et al., 2009). Whether the endogenous type I SQR may have permitted ancestral cyanobacteria to oxidize sulfide for energy, or cyanobacteria retooled this detoxifying enzyme for AP, is an open question.

The role of the second sqr operon in Geitlerinema is unknown. In addition to sulfide consumption through sqr1-based AP (Oren and Padan, 1978), Geitlerinema has a second sulfide donation site on the immediate donor side of PSI, which is not inducible (i.e., does not require protein synthesis) and does not significantly contribute to proton translocation (Shahak et al., 1987). However, this response operates at even higher concentrations of sulfide (K_m in the mM range) without saturation (Shahak et al., 1987; Arieli et al., 1991). The enzyme mediating this response to high sulfide is unknown, thus it could be encoded by this sqr2. Only 10 of the 44 cyanobacterial genomes and genomic bins examined in this study (including Geitlerinema sp. PCC 9228) have more than one version of sqr. Seven genomes in this subset, such as Synechocystis sp. PCC6803 (Nagy et al., 2014), each have two versions sqr, one of which is phylogenetically similar to sqr1, and the other a eukaryotic homolog (type II). Only three genomes, all of them Geitlerinema species, have genes phylogenetically similar to sqr2 (type VI) as well as an sqr1-like version. The sulfide physiology of Geitlerinema sp. PCC7105 is unknown, but Geitlerinema sp. BBD is a known sulfide-resistant photosynthetic cyanobacterium (Den Uyl et al., 2016). Green sulfur bacteria, thiotrophic proteobacteria, and members of Aquificaceae have multiple versions of sqr or homologs with a range of sulfide affinities. Aquifex aeolicus expresses its types I and VI sqr, similar to Geitlerinema’s sqr1 and sqr2, even when not growing thiotrophically (Marcia et al., 2010a,b). On the other hand, Chlorobium is capable of sulfide-dependent growth at mM H₂S by using its type VI sqr, and other versions are used at lower sulfide concentrations (Chan et al., 2009). Phylogenetic clustering of sqr2 with green sulfur bacterial sqrs, including one used at high sulfide levels, supports a similar role in Geitlerinema. Like sqr1. sqr2 also has a transcriptional regulatory gene arsR in the operon. These observations raise the possibility that Geitlerinema is capable of growing phototrophically on different sulfide levels through its sqr1 and sqr2. Our results also suggest that such versatility was likely achieved through a combination of evolutionary processes including vertical descent within ancestral cyanobacteria (sqr1, type I clade) as well as lateral transfer from other groups (sqr2, type VI clade). When these genes became part of the AP-cyanobacterial repertoire is uncertain.

Chromosomal examination of cyanobacterial sqr and groups 2 and 3 psbA genes underscores a potential relationship between sulfide exposure and reformulation of the water-oxidizing complex in periods of AP and/or OP. Of the evaluated cyanobacterial genomes with sqr, 15 of the genomes have sqr types I, II, or VI in close genetic proximity to low-oxygen or anaerobic psbA genes. These genomes include metagenomic bins from Middle Island Sinkhole, where sulfide-based AP has been demonstrated (Voorhies et al., 2012), two other species of Geitlerinema, four Cyanothece species, and two Leptolyngbya species, among others. Because of variable contig lengths, it is unknown if psbA1 and/or psbA2 of Geitlerinema sp. PCC 9228 are in proximity to one or both of its sqr genes. However, the small intergenetic spaces between anoxygenic/micro-oxygenic psbA varieties and sqr in the other genomes hints at potential linked transcriptional regulation of these genes. Additionally, of the cyanobacterial sqr genes, 27 were in close proximity to arsR-like transcriptional regulators. These genes and conditions of their expression are an attractive target for further studies.

Different versions of sqr in Geitlerinema sp. PCC 9228 may also be linked to trace metal metabolism and resistance. In close proximity to sqr2 there are genes for arsC. arsB. arsR, and glutathione (GST) S-transferase, and on other contigs there are genes for arsenite transporting ATP-ases arsA, another arsB, a methyltransferase similar to arsenite methylator arsM, and a chromate transporter similar to a cyanobacterial arsenite importer. These genes are involved in arsenate reduction, arsenite transport, transcriptional regulation, and mediation of arsenic resistance (Oden et al., 1994; Mukhopadhyay et al., 2002; Cameron and Pakrasi, 2010; Slyemi and Bonnefoy, 2011). Synechocystis sp. PCC6803 uses an arsR gene to regulate genomic arsBHC expression during arsenic exposure (López-Maury et al., 2003), and is able to grow in mM concentrations of arsenite and arsenate (Sánchez-Riego et al., 2014). An arsR-like transcriptional regulator is also adjacent to sqr1 in Geitlerinema (Bronstein et al., 2000), and the proximity of arsR to sqr2 suggests a similar regulatory role. Arsenic resistance genes are widely distributed among bacteria and archaea, and may be found in environments that do not have measurable arsenite (Mukhopadhyay et al., 2002; Oremland and Stolz, 2005). The hydrology of the habitat of Geitlerinema, Solar Lake, is driven primarily by seawater seepage through the sand bar, with minor meteoric input (Cohen et al., 1977a). Usual sources of arsenic, such as hydrothermal hot springs in hypersaline environments or weathering of arsenic-rich clays (Oremland et al., 2009), are absent from this system. As such, these genes may have been inherited from ancestral cyanobacteria inhabiting a metal-rich environment (Oremland et al., 2009), be used to detoxify other metals such as antimony (Nagy et al., 2014), or cope with reactive oxygen species and oxidative stress (Latifi et al., 2009; Takahashi et al., 2011).

The proximity of the sqr2 to arsenic-related genes (arsRBC, glutathione S-transferase), taken together with experimental results from related organisms and genes, hints at an unexplored potential metabolism in Geitlerinema: AP using arsenite as the electron donor, as in anoxygenic bacteria (Kulp et al., 2008; Hoeft et al., 2010; Edwardson et al., 2014). Oscillatoria-like cyanobacterial biofilms in an arsenic-rich hot spring, and Synechocystis sp. PCC 6803, can perform light-dependent arsenite oxidation (Kulp et al., 2008; Nagy et al., 2014). In Synechocystis sp. PCC 6803, the same sqr and arsR genes from its plasmid sqr operon, which enable light-dependent sulfide oxidation, are expressed during arsenite (As(III)) oxidation in the light (Nagy et al., 2014). The cyanobacterium imports arsenite through a chromate transporter (suoT, on the same plasmid operon), stores mM arsenite intracellularly without detriment, and exports excess arsenite through arsB (Nagy et al., 2014). The arrangement of Geitlerinema’s arsenic-related genes on its sqr2 operon is similar to the plasmid sqr operon of Synechocystis sp. PCC 6803. Arsenite oxidation has been linked to electron transfer to quinones (Jiang et al., 2009) and energetics support the potential of cyanobacterial plastoquinone as an arsenite oxidant (Nagy et al., 2014), but this pathway has not been explored in cyanobacteria. The co-transcription of sqr1 in Synechocystis with arsenite uptake genes during arsenite exposure (Nagy et al., 2014), and the well-characterized electron-stripping mechanism of sqr1 on sulfide (Arieli et al., 1994), hints at a role for sqr in arsenite oxidation.

Arsenite-dependent cyanobacterial AP is intriguing due to the role of arsenite-based primary production on ancient Earth. Arsenic was likely more abundant on Earth’s surface during the Archean than present (Oremland et al., 2009). In anoxic marine basins that dominated the biosphere 2.7 Ga ago, arsenite-dependent microbial autotrophy putatively cycled nitrogen, carbon, and arsenic (Sforna et al., 2014). This metabolism continues in modern hypersaline hot springs and subsurface aquifers that have elevated arsenic levels (Oremland and Stolz, 2005; Kulp et al., 2008; Hoeft et al., 2010). The genome of Geitlerinema lacks proteobacterial arsenite-oxidizing genes, and instead has genes similar to those used for light-dependent arsenite oxidation in Synechocystis sp. PCC 6803 (Nagy et al., 2014). Verifying arsenite-dependent AP in cyanobacteria, and conclusively linking the sqr2 gene in Geitlerienema to that process, would clarify the potential role of anoxygenic cyanobacteria in arsenic cycling in both modern and ancient ecosystems.

In summary, analysis of the genome of Geitlerinema sp. PCC9228 complements prior physiology studies by providing the genetic foundation for its metabolisms of nitrogen fixation, facultative OP, and sulfide-based AP. We find multiple versions of psbA, encoding a key protein for water oxidation, which may enable a sensitive response to varying conditions of light, oxygen and sulfide. Nitrogen fixation is linked to oxygen level and production via the nifI₁I₂ regulator in the nif operon and via non-oxygen producing psbA, respectively. Multiple versions of sqr likely address a range of sulfide concentrations and may also be linked to responses to metals and oxidative stress and perhaps even arsenite oxidation. Aerobic versatility encoded in the genome of Geitlerinema, coupled with diazotrophic regulation and concentration-specific sulfide responses, permit Geitlerinema to thrive in periodic sulfidic, microoxic, and poorly lit conditions of Solar Lake (Cohen et al., 1975a). Such dynamic geochemical conditions likely also challenged cyanobacteria during variable sulfide and oxygen levels of the Archean and Proterozoic (Satkoski et al., 2015; Sperling et al., 2015). This study of Geitlerinema and its unique gene assemblage addresses both the ambiguous role of Archean cyanobacteria in oxygen production/mitigation prior to the development of OP and the Great Oxidation Event, as well as the contributions of AP cyanobacteria to Proterozoic biogeochemistry. The phylogeny and diversity of genes responsible for metabolic versatility in Geitlerinema suggest a blend of genetic strategies for the anoxic early environment—such as methanogen-like modulation of nitrogen fixation and non-water oxidizing photosynthetic proteins—with post-oxidation strategies such as specific photosynthetic proteins for micro-oxic as well as oxic conditions, and different SQRs for fluctuating sulfide concentrations. Phototrophs capable of versatile AP/OP, such as Geitlerinema, would have had the advantage over organisms metabolically limited to either oxic or sulfidic conditions. Their continuous photosynthesis likely supported other microorganisms with fixed nitrogen and carbon, sulfide removal, and intermittent oxygen production. Furthermore, conditional production of oxygen at variable concentrations would have had strong influences on the structure and metabolic needs of their associated microbial communities through development of oxygen refugia and/or oases. Further research into Geitlerinema’s growth and transcriptional regulation will uncover the fine-tuned response of AP/OP cyanobacteria to changing redox conditions. In turn, we can relate the scope of these dual metabolisms and their modern physiologies to their ancestors’ impacts on ecology and geochemistry as Earth slowly and discontinuously became oxygenated.