Upon increases in extracellular osmolarity, many bacteria synthesize small organic molecules that raise the intracellular osmotic pressure. In cyanobacteria, the nature of the osmolyte correlates with the host’s osmotolerance: strains with a low salt tolerance produce sucrose; moderately halotolerant strains utilize glucosylglycerol (GG); and highly tolerant strains use glycine betaine (Hagemann, 2011).

The moderately halotolerant cyanobacterium Synechocystis sp. PCC 6803 (hereafter: Synechocystis) uses GG as its primary osmolyte (Richardson et al., 1983), but is also capable of salt-induced sucrose synthesis. GG is synthesized via a two-step pathway from central metabolites (Figure 1): in the first step, glucosylglycerol phosphate is synthesized from ADP-glucose and glycerol-3-phosphate, in a condensation reaction catalyzed by glucosylglycerol phosphate synthase (GgpS). Cleavage of the phosphate moiety is subsequently accomplished by glucosylglycerol phosphate phosphatase (GgpP). Synechocystis harbors a transporter that is used for reuptake of GG that is lost due to leakage of this osmolyte from the cytoplasm (Hagemann et al., 1997).

Another osmolyte that is frequently used by bacteria such as Escherichia coli is the disaccharide trehalose (Welsh et al., 1991). Although incapable of its synthesis, Synechocystis can take up exogenously supplied trehalose and use it as an osmoprotectant (Mikkat et al., 1997). In agreement with this it is observed that in cells to which trehalose has been added, the total concentration of GG decreases over time, suggesting that this latter molecule can be catabolized (Mikkat et al., 1997).

Some marine cyanobacteria have been demonstrated to be able to ferment a part of their osmolytes. Thus, Microcoleus chthonoplastes ferments some of its GG in the dark, re-assimilating the glucose part, while the glycerol part is excreted, or lost through leakage through the cytoplasmic membrane (Stal and Moezelaar, 1997). Yet, a metabolic pathway for GG re-assimilation has so far remained elusive in cyanobacteria (Pade and Hagemann, 2014), although recently, a dedicated GG phosphorylase was discovered in Bacillus selenitireducens (Nihira et al., 2014).

We reported elsewhere that under mild salt stress, wild-type Synechocystis cells synthesize small quantities of glycerol (Savakis et al., 2015). Here, we provide evidence supporting the notion that GG is the source of this glycerol under salt stress. In addition, we demonstrate that a previously unassigned protein, Slr1670, is directly involved in (and required for) GG re-assimilation.

Chemicals were purchased from Sigma-Aldrich, unless stated otherwise. Glucosylglycerol (GG, 51% aqueous solution) was purchased from Bitop (Germany).

When grown in presence of 200 mM NaCl, wild type Synechocystis accumulates small amounts of glycerol in the extracellular medium (Savakis et al., 2015 and Figure 4A) The transcript of glpK, encoding glycerol kinase, was reported to be upregulated under salt stress conditions (Billis et al., 2014). We expected that in the absence of other assimilation reactions, a strain deficient in glpK would produce an increased amount of glycerol when facing salt stress. Interestingly, instead, glycerol production remained unaltered in a strain in which glpK was disrupted with a chloramphenicol resistance cassette (Figures 4A,C and 5A,C). This finding suggests that glpK might not be involved in the assimilation of glycerol under the conditions tested.

To test whether instead the appearance of glycerol in the extracellular medium is dependent on the presence of intracellular GG, we analyzed a mutant strain in which ggpS is disrupted via insertion of a kanamycin resistance cassette (Du et al., 2013). This strain is sensitive to moderately high salt concentrations (∼400 mM NaCl, Supplementary Figure S1), as observed previously (20). Growth at lower concentrations of NaCl (i.e., 200 mM), however, was unaffected (Supplementary Figure S1F). In the supernatant of liquid cultures of this strain, no glycerol could be detected upon the addition of salt (Figure 4B). This indicated that GG is a precursor of glycerol under these conditions.

Significantly, within the genomic context of ggpS and glpK there is also a gene, whose transcript was shown to be upregulated under salt stress conditions (Dickson et al., 2012; Qiao et al., 2013; Billis et al., 2014). The open reading frame upstream of glpK, slr1670, has a translated length of 885 amino acids. The Slr1670 protein belongs to the GHL5 family of hypothetical glucoside hydrolases (Naumov and Stepushchenko, 2011). In order to investigate the role of Slr1670, a mutant strain was constructed in which this ORF was disrupted by a kanamycin resistance cassette (Figure 1B). Growth of the Slr1670 disruption strain was unaffected by salt (Supplementary Figures S1A,D,G,J). However, no glycerol was detectable in the extracellular medium when this strain was grown in BG11 medium supplemented with 200 mM NaCl (Figure 4D).

We then added GG to cells of the Δslr1670 strain, and analyzed the extracellular glycerol concentration in relation to the amount of glycerol formed in related deletion mutants (Figure 5). In the wild type strain, the concentration of extracellular glycerol was increased as compared to the control condition (no additional GG, Figure 5A). The ggpS disruption strain accumulated extracellular glycerol only when GG was added (Figure 5B). The amount of glycerol was lower than in the wild type strain. When GG was supplied to the ΔglpK strain, the extracellular accumulation of glycerol was increased and comparable to that of the wild type (Figure 5C). In all strains tested, the addition of extracellular GG led to an increase in growth rate in the exponential phase (Figures 5A–D). When GG was added exogenously to the slr1670 disruption mutant, no glycerol was formed (Figure 5D). These findings are indicative of Slr1670 playing a role in glycerol production from GG.

To make sure that the absence of glycerol formation could not be attributed to polar effects of the slr1670 disruption [e.g., (over-) expression of glpK from the promoter of the kanamycin resistance cassette], a mutant deficient in both slr1670 and glpK was constructed. Under salt stress, this strain failed to accumulate glycerol in the extracellular medium (Figure 5E). The slr1670 deletion strain showed increased intracellular concentrations of GG (Figure 6), corroborating the hypothesis that Slr1670 is involved in GG degradation. Taken together, these findings suggest that Slr1670 is required for the degradation of GG to glycerol.

We used genome-scale modeling, to analyze whether a possible growth advantage could be conferred by the ability to degrade GG. We started with the model published by Nogales et al. (2012), and extended it with previously published data and with the new information acquired here. Since inactivation of the ABC transporter involved in GG uptake leads to extracellular accumulation of GG (Hagemann et al., 1997; Mikkat and Hagemann, 2000), a leakage reaction from the cytoplasm to the periplasm was introduced. In this state, the model would not predict synthesis of GG when maximizing growth rate. Therefore, a sink reaction for cytoplasmic GG was introduced (Table 4; Figure 2). Since it is at this point unknown whether cleavage of GG occurs via hydrolysis or phosphorolysis (as described by Nihira et al., 2014 for the B. selenitireducens enzyme), both reactions were included.

Phosphorolysis of GG, similar to the reaction catalyzed by the enzyme of B. selenitireducens, yields phosphorylated glucose (Nihira et al., 2014). We therefore expected a clear preference for this reaction over the hydrolysis reaction. Accordingly, we simulated growth for phosphorolytic cleavage for a range of GG uptake fluxes, qGG, and divided the obtained rates by those obtained for hydrolytic cleavage (Supplementary Figure S2A). At low uptake rates (low qGG values), no difference in growth rate was predicted for either phosphorolysis or hydrolysis. Only at very high uptake rates was the phosphorolysis reaction beneficial, but even there, the predicted increase in growth rate was minor (<2%). Interestingly, the genome-scale model makes similar predictions for the benefit of glycerol utilization: for low values of qGG, no benefit for the assimilation of glycerol is predicted (Supplementary Figure S2B). Even at very high values, the increase in ratio is very small (<2%).

To estimate a physiological range for qGG, we used the glycerol production data from the strains supplemented with extracellular GG (Figure 5). The ΔggpS mutant is unable to synthesize GG, and extracellular glycerol consequently must stem from the degradation of exogenous GG exclusively (Figure 5B). We therefore used the glycerol production values from this strain to constrain qGG in the genome-scale model. Next, we simulated photo(hetero)trophic growth (μ) at 30 μE/m²/s with the resulting qGG values for wild type, ΔggpS, ΔglpK, and Δslr1670. To estimate if the degradation of GG results in increased fitness for these qGG values, we divided the growth rate of the respective strains by the growth rate of the Δslr1670 mutant and compared these values to the experimentally determined data (Table 5). In the exponential phase (days 0–4) there is excellent agreement between simulation and experiment. Between day 4 and day 6 and between day 8 and day 10, the predicted increase is lower than in the experiment. As the culture increases in density, the absolute value of photons that a single cell perceives decreases. This results in slower (linear) growth (due to light limitation). Under the assumption that the contribution of GG utilization to growth remains constant, the difference in growth (expressed as the ratio) between a strain that utilizes GG, and one that cannot, will therefore be amplified under low light conditions. Since we used a constant photon uptake rate for the simulations, it is expected that at later time points, the effect of GG utilization is underestimated. In the strains that do have a functional slr1670, on day 8, there is a drop in optical density (Figures 5A–C), which results in negative values for the experimentally determined growth rates. Since the model can only predict growth, a comparison between the experimental and simulated ratios is not meaningful for this and the adjacent intervals. The cause of this transient drop in OD is at present unknown. A possible explanation is that the degradation of GG, and utilization of the glucose part thereof, leads to a drop in intracellular osmotic pressure. Transient efflux of water could then reduce the cell volume and hence, the optical density.

Until now, Slr1670 is annotated as a hypothetical protein, so we decided to study its phylogenetic distribution. A PSI BLAST of the translated sequence of Slr1670 yielded around 90 homologs, which are distributed among the cyanobacteria, α-proteobacteria and Archaea (see Figure 3 for a condensed tree, including only the cyanobacterial species and Supplementary Figure S3 for the full phylogenetic tree). In many of the cyanobacterial strains, also homologs of ggpS, ggpP, and glpK could be identified (Figure 3). Notably, a homolog was also found in M. chthonoplastes, a strain that has been shown to ferment GG under dark conditions, thereby utilizing the glucose moiety, but excreting glycerol (Moezelaar et al., 1996).

Production of one GG molecule requires fixation of 9 CO₂ molecules. From a cellular/physiological point of view, GG is an energetically costly compound with often transient use only. It is therefore conceivable that systems have evolved to reduce energy losses related to the synthesis of this osmolyte. The GgtABCD system takes up GG that is lost from the cells due to leakage from the cellular cytoplasm; inactivation of this system leads to pronounced extracellular accumulation of GG (Hagemann et al., 1997; Mikkat and Hagemann, 2000). Upon transition from a high-salt to a low-salt environment, osmoprotective compounds are no longer needed to maintain turgor pressure, so cells able to salvage the nine carbon atoms of GG are at an advantage. In line with this, the adjusted genome scale model of Synechocystis predicted an increase in growth rate when extracellular GG was made available (Table 5).

Marine cyanobacteria may ferment a portion of their osmoprotectants (Stal and Moezelaar, 1997). For example, M. chthonoplastes ferments GG, using only the glucose moiety and excreting the glycerol (Moezelaar et al., 1996). In M. chthonoplastes, GG was used as a substrate for fermentation during the night. Interestingly, a gene homologous to slr1670 was also found in the genome of M. chthonoplastes (Figure 3), suggesting that the corresponding protein is involved in the degradation of GG in this strain.

At this moment, it is not known if the degradation of GG, observed under continuous illumination in Synechocystis, a freshwater organism, serves a regulatory purpose or is the result of a constitutive reaction that becomes more meaningful in the dark. During growth with constant incident illumination, the average amount of light per cell decreases. The individual cells, however, perceive something different. At a given moment, the cells located close to the light source receive light at a high intensity, while a cell in the center of the culture receives fewer photons or none at all. Increases in culture density therefore may lead to increased periods of darkness from the perspective of an individual cell. This is corroborated by the observation that more glycerol is excreted per cell in the late versus the early phase of the experiment (Figures 4 and 5). This notion might explain why glucose recycling becomes active during the day.

Cellular release of glycerol by both Synechocystis and M. chthonoplastes is in agreement with the prediction from the genome scale model, i.e., that glycerol utilization does not lead to increased growth at physiological levels of the GG assimilation reaction (Supplementary Figure S2B). The mutant strains deficient in degradation of GG are valuable tools to study the kinetics of GG synthesis.

Under industrial production settings, Synechocystis may be exposed to variations in osmotic stress. It is important to be able to simulate the behavior of this organism under such condition in order to facilitate its usage in the direct conversion of CO₂ to products of interest in a sustainable fashion (dos Santos et al., 2014). The addition in the model of the ability to synthesize, transport and degrade GG is a necessary step toward the accurate simulation of the growth of Synechocystis in such conditions. Additionally, disruption of slr1670 allows the construction of strains with increased levels of GG, an interesting commodity chemical (Tan et al., 2015).

We have demonstrated a function for a previously non-annotated protein that has turned out to be required for the re-assimilation of GG. The protein shows homology to glucoside hydrolases and is found in a wide range of cyanobacteria, where its presence correlates with the presence of genes required for GG synthesis. Furthermore, homologs of this protein are also found in the α-proteobacteria and in the domain of the Archaea (Supplementary Figure S3). Strikingly, halotolerant organisms are overrepresented in these two classes of organisms, particularly among the Archaea. Often, homologs obtained had previously been annotated as alpha-amylases. These annotations, however, are purely based on homology and are not supported by experimental evidence. It is also of interest to note that the genomes of some cyanobacterial species encode multiple reading frames that exhibit sequence similarity with Slr1670.

Genome-scale modeling revealed that the increase in growth rate caused by the utilization of the glucose part of GG outweighs the use of the glycerol part by far. This is in line with observations in M. chthonoplastes, in which only the glucose part of GG is utilized by the cells. Studies with the purified enzyme will help to elucidate the kinetic properties of Slr1670 and shed light on the mechanism of cleavage of GG.

PS and XT conceived the study, designed and carried out the experimental plan, and wrote the paper. CQ and KS implemented the experimental plan. XL, KH, and FBS conceived and supervised the study and wrote the paper.

KH is advisor to Photanol B.V. The remaining 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.