Cyanobacterial photosynthesis under sulfidic conditions: insights from the isolate Leptolyngbya sp. strain hensonii

We report the isolation of a pinnacle-forming cyanobacterium isolated from a microbial mat covering the sediment surface at Little Salt Spring—a flooded sinkhole in Florida with a perennially microoxic and sulfidic water column. The draft genome of the isolate encodes all of the enzymatic machinery necessary for both oxygenic and anoxygenic photosynthesis, as well as genes for methylating hopanoids at the C-2 position. The physiological response of the isolate to H2S is complex: (i) no induction time is necessary for anoxygenic photosynthesis; (ii) rates of anoxygenic photosynthesis are regulated by both H2S and irradiance; (iii) O2 production is inhibited by H2S concentrations as low as 1 μM and the recovery rate of oxygenic photosynthesis is dependent on irradiance; (iv) under the optimal light conditions for oxygenic photosynthesis, rates of anoxygenic photosynthesis are nearly double those of oxygenic photosynthesis. We hypothesize that the specific adaptation mechanisms of the isolate to H2S emerged from a close spatial interaction with sulfate-reducing bacteria. The new isolate, Leptolyngbya sp. strain hensonii, is not closely related to other well-characterized Cyanobacteria that can perform anoxygenic photosynthesis, which further highlights the need to characterize the diversity and biogeography of metabolically versatile Cyanobacteria. The isolate will be an ideal model organism for exploring the adaptation of Cyanobacteria to sulfidic conditions.


Enrichment and isolation
Samples of red pinnacle mat were homogenized and a small aliquot (~ 500 µL) was added to BG11 media (Rippka et al., 1979) supplemented with 25 mM HEPES (B-HEPES) and adjusted to pH 7.2. Enrichment cultures were maintained in 60-mL of liquid media in 125-mL conical flasks at 100 RPM at 28 ºC under either a day-night cycle or continuously illuminated with 100 µmol photons m -2 s -1 under cool white fluorescent lamps. An axenic culture was achieved using a dilution series in liquid media where the highest dilution which showed growth was taken as inoculum for the next dilution. Light microscopy was performed periodically to visually examine enrichments for purity. The dilution to extinction strategy was continued until light microscopy indicated the presence of a single morphotype and sequencing of the 16S rRNA gene returned a single phylotype. Growth of the isolate was monitored with chlorophyll a concentration determined spectrophotometrically using the absorption at 665 nm of a methanol extract and an extinction coefficient of 0.075 ml µg −1 (made from a filtered 2-ml culture subsample) (De Marsac and Houmard, 1988) or protein concentration using the Bradford assay (Bradford, 1976) with bovine serum albumin (Sigma-Aldrich, St. Louis, MO) as the standard.

Nucleic acid analyses
Samples of biofilm (~ 1.5 mL) were harvested by centrifugation, the excess media removed by decanting, and cell pellets frozen immediately (-20 °C) or subjected to nucleic acid extraction. Genomic DNA was extracted as previously described (Boyd et al., 2007). Quality of extracted DNA was assessed on an agarose gel (1%) using the HiLo DNA Marker (Bionexus, Oakland, CA) and visualized by ethidium bromide staining and using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, Delaware). To check for purity/contaminants, 16S small subunit RNA genes were amplified with bacterial domain primers 27F and 1492R were used (Lane, 1991) as described previously (Hamilton et al., submitted). Reactions were performed in triplicate, purified using a QIAquick PCR Purification Kit (Qiagen, Valencia, CA) and sequenced at the Genomics Core Facility of the Huck Institutes of the Life Sciences at Penn State University. Sequences were assembled and manually checked using using Bio-Edit (v7.2.5), and checked for chimeras using CHIMERA_CHECK (Cole et al., 2003). Putative chimeras were excluded from subsequent analyses. A single 16S rRNA gene sequence was recovered indicating the culture was pure.

Genomic sequencing
Double-stranded gDNA was quantified by Qubit (Life Technologies, Carlsbad, CA) and 100 ngs of DNA was fragmented for 7 minutes at 37 ºC using the Ion Shear Reagent according to the Ion Xpress™ Plus gDNA and Amplicon Library Preparation Users Guide (Life Technologies, Carlsbad, CA). Fragments ! 2 were purified using AMPure XP Reagent (Beckman Coulter, Brea, CA) and checked for size on a Bioanalyzer DNA High Sensitivity chip (Agilent Technologies, Santa Clara, CA). Adaptors were ligated to the fragments and DNA was nicked, repaired and then purified. A 315 bp library was size-selected on a Pippin (Sage Science, Beverly MA) using the "tight setting". Following purification, the library was checked again on the Bioanalyzer and then amplified on beads in an emulsion PCR using the Ion Xpress Template Kit and 200 base sequencing was performed using a 316 chip on the Ion PGM machine as per the Users Guide (Life Technologies, Carlsbad, CA) at the Genomics Core Facility of the Huck Institutes of the Life Sciences at Penn State University.

Phylogenetic analysis
For gene-specific phylogenetic analyses, full-length sequences were identified in the genome using functional annotation and BLASTX, translated, and verified by BLASTP. Reference datasets were populated by detecting homologs in IMG genomic databases by BLASTP (Altschul et al., 1997). Protein sequences were aligned with MUSCLE (Edgar, 2004) and redundancy in the alignments was reduced through the Decrease Redundancy Program (http://web.expasy.org/decrease_redundancy/). Maximum likelihood trees were constructed using PhyML (Guindon et al., 2010) with the LG+gamma model, four gamma rate categories, ten random starting trees, NNI branch swapping, and substitution parameters estimated from the data. The resulting trees were viewed and edited using iTOL (http://itol.embl.de/) (Letunic and Bork, 2016

MODEL FOR THE KINETIC REGULATION OF OXYGENIC AND ANOXYGENIC PHOTOSYNTHESIS
Leptolyngbya sp. strain hensonii shows a complex response to H2S: (I) rates of anoxygenic photosynthesis (AP) are regulated by both H2S and irradiance (Fig. 5); (II) O2 production is entirely inhibited by minute H2S concentrations (Fig. 4); (III) the recovery rate of oxygenic photosynthesis (OP) after inhibition is dependent on irradiance (Fig. 6); and (IV) AP rates can be double those of OP ( Fig. 3; Fig. 5). To constrain the range of possible biochemical mechanisms shaping this response, we constructed a simplified model of the electron transport reactions involved in oxygenic and anoxygenic photosynthesis.
The model described here represents a fusion and further developed version of models described in Klatt et al., 2015a, and2015b. Using this model we show that the inhibition and recovery of OP can be explained based on the interaction of H2S with an intermediate of the oxygen evolving complex (OEC), the generation of which is light dependent. Also, we show that the regulation of AP is substantially more complex than in previously studied cyanobacterial strains and cannot be explained by just considering the properties of SQR, i.e., its affinity for oxidized PQ and H2S. Instead, we suggest that H2S affects reactions downstream of PSI and we present two independent hypotheses concerning the light-dependency of the kinetics of H2S oxidation: (a) light-dependent activity of SQR and (b) multiple H2S oxidizing enzymes.

Oxygenic photosynthetic electron transport reactions
Cyanobacteria perform OP via electron transport reactions that are often represented as the Z-scheme ( Fig. S3) and that have been intensively reviewed (reviewed e.g. in (Vermaas 2001;Govindjee et al., 2010;Shevela et al., 2013;Roach and Krieger-Liszkay, 2014;Johnson 2016)). In brief, photosynthesis is initiated by the absorption of photons by pigments associated with PSII and PSI. The light energy is absorbed by accessory (also: antennae) pigments, such as phycocyanin, phycoerithrin, and other chlorophylls, is then transferred in the form of excitation energy to the reaction center (RC) chlorophyll a (Chl a). Via various intermediate redox reactions the excited state reaction center Chls drive the oxidation of the external electron donor H2O and the reduction of redox carriers NADP + . During electron transport a proton motive force (pmf) is generated, which drives the formation of ATP. Thus, during these "light reactions" of this linearly organized electron transport chain the excitation energy is converted into chemical energy in the form of reduced electron carriers and ATP. The NADPH and ATP generated during these electron transport chain reactions are driving CO2 fixation in the "dark" reactions of the Calvin cycle. The individual reactions involved in oxygenic photosynthetic electron transport are described in greater detail below.
As cyanobacteria are often exposed to fluctuating environmental conditions, the photosynthetic electron transport and its regulation are complex (reviewed e.g. in (Peschek, 1999;Vermaas, 2001;Mullineaux 2014a;Nagarajan et al., 2015). There are multiple possibilities for alternative electron flow, ! 4 i.e., "branching" of the main electron transport highway. The "valves" of the linearly organized photosynthetic electron transport chain become important regulators when otherwise excessive oxidation or reduction of the transport chain components would occur.
Most prominently, the PQ pool, the cytochrome b6f complex, and the soluble electron carriers (plastocyanin or cytochrome c553) are also shared with the respiratory pathway (Fig. S3). PQ in the thylakoid membranes can receive electrons from succinate dehydrogenase (SDH). Reduced PQ, plastoquinol, can directly donate electrons to terminal oxidases, such as the plastoquinol oxidase (PTOX), the alternative respiratory terminal oxidase (ARTO) and cytochrome bd-type quinol oxidase (Cyd) (Lea-Smith et al., 2016). These oxidases couple the oxidation of PQ to the reduction of O2. In addition, the soluble cytochromes can donate electrons to a terminal oxidase, aa3-type cytochrome-c oxidase complex, instead of PSI. It is important to note for the discussion below that genes encoding both a cytochrome c oxidase and a bd-type quinol oxidase were identified in the strain hensonii genome. PQ reduction by SDH and PQ (or cytochrome) oxidation via terminal oxidases can likely occur simultaneously with photosynthetic electron transport, with O2 reduction representing an electron "dump". These reactions are likely important for cyanobacteria to maintain redox balance and prevent cell damage by photoinhibition but can also be dangerous (Murata et al., 2012). Cyclic electron transport can also occur via electron transfer from Ferredoxin to PQ. Furthermore, PSI and PSII have several other alternative electron acceptors, such as Flavodiiron proteins (Flv1-4) and a bidirectional [NiFe] hydrogenase (Hox) (Zhang et al., 2012;Gutekunst et al., 2014;Mullineaux, 2014b).
While possibly representing important regulatory functions to maintain redox balance, we assume here for simplicity that the rates of cyclic electron transport and other alternative electron flow reactions are insignificant compared to the photosynthetic electron mainstream reactions. Our methods do not allow for a quantification of the rates of these reactions and, as demonstrated below, their consideration is not needed to explain our data. We therefore limit our simplified model to the main components of the electron transport chain. Furthermore, we only concentrate on the reaction rates that can become the kinetic bottleneck for photosynthetic electron transport.

Formulation of Rate laws for oxygenic photosynthesis
In our model we assume that during non-saturating light conditions, light energy harvested in PSI or PSII limits OP, as manifest in the linear increase of OP with light in PI-curves ( Fig. 7) (Klatt et al., 2015b).
Thus, the light energy harvested by accessory pigments that has migrated to the Chl a in the reaction center of PSII determines the rate (kE) of generation of an excited catalytic Chl a dimer in PSII (PSII*).
This can be formulated as where PSII is ground state Chl a in PSII and fII is an absorbance cross-section factor that describes the efficiency of conversion of the externally available photon flux (E) into a volumetric rate of excitation.
Upon the generation of an excited Chl a in PSII, charge separation between this Chl and primary electron ! 5 acceptor pheophytin (Pheo) occurs. The term charge separation describes the formation of a radical pair, P680•+ and Pheo•-in PSII. This is the initiation of the electron transport reactions. Pheo rapidly reduces a bound plastoquinone (QA), which donates electrons to a next redox carrier, another plastoquinone (QB), which joins the mobile plastoquinone pool (PQ). While the initial reactions are extremely fast and occur directly in PSII RCs, PQ reduction and mobilization, and the following re-oxidation by cyt b6f are comparatively slow, which is related to the diffusion time of the PQ molecules in the thylakoid membranes (Whitmarsh and Govindjee, 1999). As a result, the redox state of the PQ pool is highly sensitive to imbalances between energy harvested in PSI and II, and to the rate saturation downstream of PSI. In other words, PQ is an important indicator of the overall redox state of the photosynthetic electron transport chain. Importantly, the availability of oxidized PQ can become the rate limiting factor for OP.
In our model the rate of electron transport initialization is, however, not only dependent on the availability of oxidized PQ but also to the availability of a functional oxygen evolving complex (OEC) that is not inhibited by H2S and can mediate the oxidation of H2O to O2. As mentioned above, charge separation leads to the formation of an oxidized RC Chl a (P680+). Electrons are rapidly transferred from a tyrosine residue of the OEC, which is in turn reduced by the Mn4OxCa cluster of the OEC. Overall, the rate of PQ reduction by PSII (kOP) can thus be described as following Michaelis-Menten kinetics according to where PSII* is the excited catalytic Chl a dimer in PSII, OEC is non-inhibited oxygen evolving complex and PQox is the oxidized part of the plastoquinone pool. vOP is the maximum rate and KOEC and KPSII are the reaction constants describing the apparent affinity of PSII* towards OEC and oxidized PQ.
To oxidize two water molecules and generate one O2 molecule, four oxidizing equivalents in the Mn4OxCa cluster have to be generated ultimately driven by 4 photons harvested by PSII. The exact reactions occurring in the OEC are still to be fully described (Messinger and Renger, 2008;Jablonsky and Lazar, 2008). Clearly, the OEC undergoes several transitional stages (initially described as S states; (Kok et al., 1970)) before release of an O2 molecule. In our model, we use "OECox" to summarize multiple  (Nishiyama et al., 2006). The rate of PSII degradation will therefore depend on the availability of the excited catalytic Chl a dimer in PSII (PSII*) and irradiance (E) according to KD1 was chosen such that the rate begins to saturate at high light intensities because we also assumed that the efficiency of photoinhibition is light-dependent (Keren et al., 1997). Degraded subunits of PSII are repaired at a rate described by where PSIId is non-active PSII. This reaction leads to the replenishment of active ground-state PSII.
The successful initialization of electron transport yields reduced PQ according to Eq. S2.
Oxidation of PQ occurs via the cyt b6f complex and plastocyanin (or cytochrome c553), and is driven by light energy harvested in PSI. From PSI electrons are transferred via a series of fast intermediate redox reactions to ferredoxin, which, via ferredoxin-NADP + oxidoreductase (FNR), can reduce NADP + . The rate of PQ oxidation can thus be described as where fI is an absorbance cross-section factor of PSI. The factor β introduces excitation energy transfer from PSII to PSI dependent on redox state of the PQ pool. Namely, with increasing reduced PQ (PQred) in the total PQ pool (PQtot) an increasing fraction β of light energy harvested in PSII is transferred to PSI, which represents a mechanism to prevent an electron "traffic jam".
The process described by Eq. S8 yields oxidized PQ that is available again for reduction by PSII and NADPH. The redox carrier NADPH is used in the Calvin cycle for CO2 reduction. The rate of CO2 fixation coupled to NADPH oxidation, which depends on the maximum rate of CO2 reduction, and the availability of NADPH, which can be formulated as ! 7 ! . (S10) The model for oxygenic photosynthetic electron transport is solved by finding steady state solutions to six differential equations: describes the rate of change in the fraction of RC Chl a that is in excited state. PSII* is generated at the rate kE (Eq. S1) and decreases due to degradation of PSII* and by OEC reduction at rates kD and kOP, respectively. The ground state fraction of PSII is generated at a rate described as as a result of OEC oxidation coupled to PQ reduction (kOP) and repair of degraded PSII (kR). The rate of PSII decrease by excitation is kE. Importantly, the sum of all fractions of photosystem II are constant in our model, which can be written as PSIItot = PSII + PSII* + PSIId.
The fraction of unstable intermediately oxidized OEC (OECox) in the total OEC pool, is increasing during re-establishment of ground state PSII and PQ reduction (kOP) and "de-inhibition" by H2S (kS2), and decreasing due to inhibition (kS1) and H2O oxidation (kO2) The rate of change of the oxidized fraction of OEC (OECox) is thus described as while the rate of change of the fraction of OEC in the total OEC pool follows which describes that OEC is a result of H2O oxidation (kOP), and regeneration of PSII and reduction of PQ (kOP). Similarly, the differential equation which also applies to the total NADP + /NADPH pool (NADPtot), i.e., ! 8 NADPtot = NADP + + NADPH. (S19) In our model, NADPH is generated during PQ oxidation (ktot) and consumed by CO2 fixation (kCO2), which is described as By setting dPSII*/dt = 0, dPSII/dt = 0, OECox/dt = 0, dOEC/dt = 0, dPQox = 0 and dNADPH/dt = 0, the steady-state redox state of the PSII, OEC, PQ and NADP pool and the corresponding rate of oxygenic photosynthetic electron transport can be calculated using Eqs. S1 -S11 for any given values of [H2S] and E.
Our data was fitted using a numerical implementation of the model into R with the deSolve package (Soetaert et al., 2010). When testing reasonable values for the maximum rates vx, affinity constants Kx and for the factors fI and fII in Eqs. S1 -S10 we considered that at saturating light intensities, rates of OP stop increasing linearly with light ( Fig. 7) and rates of CO2 fixation are expected to become the bottleneck of OP rates. Therefore, KCO2 and vCO2 in Eq. S10 had to be set substantially higher than all other maximum rates and affinity constants to achieve that kCO2 limits the overall electron transport at high light only. We also considered that in most cyanobacteria PSI RCs are more abundant in the thylakoid membrane than PSIIs to prevent that ktot becomes the bottleneck for electron transport because an electron transport "traffic jam" can cause photodamage to PSII, as described above. We therefore set fI>fII, which revealed very good agreement with the experimental data (see dashed line in Fig. 7).
In the formulation of OECox inhibition by H2S (Eq. S4), the rate of inhibition will be dependent on light because the abundance of OECox is light-dependent (see Eq. S1 and S2). Klatt et al. (2015b) have indeed observed pronounced light-dependent inhibition kinetics in their studied Planktothrix strain. We could not observe similar kinetics in Leptolyngbya sp. strain hensonii -H2S instantaneously inhibited OP at very low concentrations. Therefore, κ was chosen to be substantially higher than in the model for the cyanobacteria previously described in Klatt et al. (2015b).
After exposure to H2S, oxygenic photosynthetic rates only started to recover after a fixed time frame of ~30 min. This could be explained by considering in our model that σ<<< κ in Eqs. S4 and S5, which implies that any available OECox is rapidly inhibited by H2S, while the back reaction is so slow that the inhibited state of the OEC dominates for an extended time frame even after removal of externally available H2S. Furthermore, the recovery rate of oxygenic photosynthesis from H2S inhibition was lightdependent in strain hensonii (Fig. 6). Inhibition of photosynthesis by H2S on the donor side of PSII, is expected to cause photodamage of PSII (Jegerschoeld et al., 1990) because the excitation energy cannot be converted into chemical energy in the form of electron transport. We adjusted constants in our model such that an active OEC is only slowly re-generated even after depletion of H2S. Thus, excited/oxidized RC Chl a (PSII*) is more abundant than without "H2S-history" and rates are highly dependent on light (E) excuse light determines degradation rates (Eq. S6). The output of our model was in remarkable ! 9 agreement with the experimental data (see black lines in Fig. 6). We therefore conclude that the lightdependent recovery rates can be explained based on the intertwined kinetics of OEC inhibition by H2S and photoinhibition. i.e., light-dependent degradation of D1 in PSII.

Anoxygenic electron transport reactions
The model considers that in photosynthetically versatile cyanobacteria, electrons derived from H2S oxidation enter the electron transport chain at the level of the PQ pool via SQR (Fig. S3). Thus, the electron transport pathways of AP and OP intersect in the PQ pool. Cyanobacteria therefore use all redox active compounds downstream of PQ, towards NADP + and ultimately to CO2, for the transport of both H2S-and H2O-derived electrons. It is thus a crucial assumption in our model that the kinetics of photosynthetic electron transport beyond PQ are independent of photosynthetic mode unless direct modifications by H2S occur.

Formulations of rate laws for anoxygenic photosynthetic electron transport
The rate of H2S oxidation is expected to be dependent on the affinity of SQR to H2S and to PQ, which can be formulated as according to Klatt et al. (2015a). The differential equation S17 thus changes to The rate of AP depends on the availability of H2S and redox status of the PQ pool.
In our model all rates are in µM electrons s -1 . ultimately governing also the rate of H2S oxidation. We observed that electron transport rates of AP were double those of OP at high light conditions ( Fig. 5; Fig. 7). High H2S concentration, however, caused a decrease of rates (Fig. 5). We therefore reformulated Eq. S10 to where ! , which describes the enhancement of maximum rates of CO2 fixation with H2S, and where The numerical implementation of this model gave excellent results for the enhancement and inhibition of AP at higher E. However, we could not explain the light dependent slope of rates of AP in Fig. 5. This is because our model predicts that rates of AP are increasing with [H2S] until a lightdependent maximum is reached, i.e., until PSI turns into the bottleneck for electron transport. Below this maximum, light does not affect rates of AP (Klatt et al., 2015a). We could only achieve an agreement of our model with the experimental data by using either of two assumptions: (i) Light modulates the activity of SQR by increasing vmax in Eq. S21 and (ii) there are two sulfide oxidizing enzymes that donate electrons into the electron transport chain at different levels. In the following, we introduce two different versions of the model based on these two assumptions.

Light-dependent vmax
The light dependency of rates of AP at low H2S concentrations could be explained by light-dependent activity of SQR. This could be, for instance, due to light-dependent expression of SQR or even due to multiple SQRs, the relative abundance of which is controlled by light. In any case, this can be modeled by introducing light-dependent changes of vmax in Eq. S21 as where α increases with E according to The changes of H2S oxidation rates in response to light were, however, instantaneous, which means that hensonii would possess a mechanism to continuously adjust the activity of SQR according to momentary irradiance. We considered this unlikely and thus considered an alternative explanation: the presence of a second, unidentified, sulfide oxidizing enzyme ("USO"). This assumption could explain our data as described below.

An additional sulfide oxidizing enzyme ("USO")
Assuming that a second sulfide oxidizing enzyme ("USO"), such as flavocytochrome c oxidase, exists, we implemented H2S oxidation by this enzyme into our model as where "ECox" is any component of the electron transport chain between PQ and PSII. ECox thus receives electrons from both PQ and "USO". The rate of PQ oxidation (Eq. S8) thus changes to ! and is not directly coupled to reduction of NADP + anymore. NADP + reduction via PSI in our model is instead coupled to EC oxidation according to ! , where β is defined by Eq. S9.
Consequently, also the differential equation describing the rate of change of the oxidized fraction of the PQ pool (Eq. S22) had to be altered to ! .
We additionally introduced the rate of change of the oxidized fraction of the EC pool as ! and when finding a solution the sum of the total EC pool was constant as "EC"tot = "EC"ox + "EC"red.
The relative contribution of the two sulfide oxidizing enzymes to the overall sulfide oxidation activity will depend on their affinity to H2S and the redox state of the respective electron acceptor in the transport chain. Assuming that "USO" donates electrons downstream of PQ it will cause its electron acceptor to become more reduced. This electron acceptor EC is also the electron acceptor for PQ. The activity of "USO" thus affects the oxidation state of PQ and thus SQR activity. At low light conditions competition for the electron transport chain takes effect. This competition scenario can occur when setting ! 12 KPQ and KC1 > KN and KC2, which will lead to accumulation of reduced PQ at increasing H2S concentrations. Fig. S4A shows that consequently rates of H2S oxidation by SQR abruptly stop increasing at very low H2S. This effect is most pronounced when setting the affinity constant of "USO" (KM2) lower than that of SQR (KM). At higher light, rates of overall electron transport through PSI are sufficiently high to keep both the reduced fraction of PQ and "EC" low enough to circumvent that PQ oxidation is limiting SQR activity at a low H2S concentration (Fig. S4B). Then the rates of sulfide oxidation are strictly determined by the affinities to H2S and not to the electron acceptors PQ and EC.
Our data set does not enable evaluation of both hypotheses. Regardless, the main conclusion of our modeling effort is that anoxygenic photosynthesis in Leptolyngbya sp. strain hensonii is controlled by substantially more complex mechanisms than expected based on previous studies, e.g. in Planktothrix FS39 described by Klatt et al. (2015a). Most prominently, rates of AP cannot be explained by assuming a steady pool of one single sulfide oxidizing enzyme.
! 13 Figure S1. A). Schematic diagram of the experimental setup used in this study to measure rates of oxygenic and anoxygenic photosynthesis using microsensors. Cyanobacterial biofilms grown on glass fiber filters (a) were fixed on polyester fibrous web (b) that separated a top and bottom reservoir of media.