Carbon signaling protein SbtB possesses atypical redox-regulated apyrase activity to facilitate regulation of bicarbonate transporter SbtA

Significance

Life on Earth relies on CO₂ fixation via the Calvin–Benson–Bassham cycle, which evolved first in cyanobacteria. Recently, we identified the protein SbtB, which is ubiquitously distributed in cyanobacteria, as a carbon-sensing module, controlling the HCO₃⁻ transporter SbtA and glycogen metabolism. SbtB senses various adenine nucleotides and moreover possesses the R-loop, a C-terminal hairpin loop that is implied to be involved in redox sensing. Here, we report that SbtB exhibits an unusual ATP/ADP diphosphohydrolase activity, which is modulated by the redox state of the R-loop. Thereby, SbtB is able to switch between different adenyl nucleotide-bound states in response to the cellular redox state, which depends mainly on day/night changes, highlighting SbtB as a central switch in cyanobacteria.

Abstract

The PII superfamily consists of widespread signal transduction proteins found in all domains of life. In addition to canonical PII proteins involved in C/N sensing, structurally similar PII-like proteins evolved to fulfill diverse, yet poorly understood cellular functions. In cyanobacteria, the bicarbonate transporter SbtA is co-transcribed with the conserved PII-like protein, SbtB, to augment intracellular inorganic carbon levels for efficient CO₂ fixation. We identified SbtB as a sensor of various adenine nucleotides including the second messenger nucleotides cyclic AMP (cAMP) and c-di-AMP. Moreover, many SbtB proteins possess a C-terminal extension with a disulfide bridge of potential redox-regulatory function, which we call R-loop. Here, we reveal an unusual ATP/ADP apyrase (diphosphohydrolase) activity of SbtB that is controlled by the R-loop. We followed the sequence of hydrolysis reactions from ATP over ADP to AMP in crystallographic snapshots and unravel the structural mechanism by which changes of the R-loop redox state modulate apyrase activity. We further gathered evidence that this redox state is controlled by thioredoxin, suggesting that it is generally linked to cellular metabolism, which is supported by physiological alterations in site-specific mutants of the SbtB protein. Finally, we present a refined model of how SbtB regulates SbtA activity, in which both the apyrase activity and its redox regulation play a central role. This highlights SbtB as a central switch point in cyanobacterial cell physiology, integrating not only signals from the energy state (adenyl-nucleotide binding) and the carbon supply via cAMP binding but also from the day/night status reported by the C-terminal redox switch.

The proteins of the PII signal transduction superfamily are widespread in all domains of life, representing one of the most ancient and largest signaling protein families in nature (1). These proteins are characterized by their highly conserved trimeric structure, consisting of a triangular core of β-sheets from the ferredoxin-like fold of the three subunits (2–6). These subunits have three characteristic loop regions (T-, B-, and C-loops) (1, 3–9), which are located near the intersubunit clefts and play a major role in ligand binding and intramolecular signaling. Despite the highly conserved structure, their amino acid sequence conservation is often low, implying that the PII superfamily also comprises members that are involved in the regulation of cellular activities that differ markedly from those controlled by canonical PII proteins (1–5). In cyanobacteria, in addition to the canonical PII protein involved in N/C sensing (1, 10), the PII-like protein SbtB evolved to regulate the cyanobacterial carbon-concentrating mechanism (CCM) (3, 11).

Cyanobacteria use the CCM to cope with limiting CO₂ levels, thus augmenting intracellular levels of inorganic carbon (hereinafter C_i; referring to bicarbonate and CO₂) and providing the major carbon-fixing enzyme, ribulose 1,5-bisphosphate carboxylase/oxygenase (RubisCO), with CO₂. This ensures efficient carboxylation of ribulose 1,5-bisphosphate and represses RubisCO’s oxygenation reaction (10). Cyanobacterial CCM comprises several CO₂ and HCO₃⁻ uptake systems. The high-affinity sodium-dependent bicarbonate transporter SbtA is one component of the CCM system and is highly expressed under C_i limitation together with its downstream gene encoding for SbtB. Previous works demonstrated that SbtA is the primary target of SbtB (3, 8, 9, 12). In our previous work, we provided structural, biochemical, and physiological characterization of this unique PII-like protein from Synechocystis (ScSbtB) (3, 11, 13). The binding properties of ScSbtB turned out to be unique among the so-far characterized members of the PII superfamily, as it can bind a variety of adenine nucleotides (ATP, ADP, and AMP), including the second messenger cyclic AMP (cAMP). Recently, we additionally reported the binding of the second messenger cyclic di-AMP (c-di-AMP) to ScSbtB, which plays a key role in the regulation of diurnal and carbon metabolism via controlling glycogen synthesis (11, 13). We could demonstrate that c-di-AMP binding promoted the interaction of SbtB with the glycogen-branching enzyme GlgB, identifying the latter as another important target of SbtB (13).

Among the different adenyl nucleotides, ScSbtB exhibited the highest affinity for cAMP and c-di-AMP, followed by ADP, ATP, and then AMP. All of them bind to the canonical binding sites at the intersubunit clefts, and the binding modes of AMP (PDB: 5O3R), cAMP (PDB: 5ORQ), and c-di-AMP (PDB: 7OBJ) were disclosed in crystal structures. Intriguingly, although AMP and cAMP were found to convey opposite signals, a comparison of the ScSbtB:AMP and ScSbtB:cAMP complexes did not reveal any conformational differences on the flexible surface-exposed T-loop (3), the canonical protein interaction module of PII proteins. In contrast, the ScSbtB:c-di-AMP complex revealed a structural rearrangement of the T-loop in response to c-di-AMP binding (13).

It is generally evident that SbtB exerts its regulatory functions on SbtA and GlgB through the differential binding of these adenine nucleotide effector molecules, which may induce conformational changes on the T-loop, in analogy to canonical PII proteins (1, 10). This was underlined by recent studies that revealed the structure of the SbtB:SbtA complex (8, 9), showing a symmetric binding of the trimeric SbtB protein to a trimeric SbtA structure in which the T-loops of SbtB lock partially the C_i channels of SbtA. This complex is stabilized by the AMP-bound state of SbtB, but it is sterically incompatible with cAMP and ATP binding (8, 9). However, the structural details of how SbtB regulates other targets, like GlgB, remain elusive.

All available data imply that SbtB regulates the activity of its targets in response to the concentrations of intracellular adenyl-nucleotides and the redox or photosynthetic state of the cell (i.e., day–night transition). In this context, a further striking feature of ScSbtB is a C-terminal extension that contains a highly conserved CGPxGC motif, which is widespread among SbtB proteins (3). This extension forms a small hairpin loop structure, in which a disulfide bond is found between the two cysteines C105 and C110 (3). We hypothesized that this extension might be involved in a redox-sensory function that may read out the redox or photosynthetic state of the cell and integrate this information into the regulation of its targets; in the following, we will call this extension R-loop (for Redox-regulated loop). In this study, we used structural and biochemical approaches to reveal an intricate interplay between nucleotide binding and the redox state of the R-loop. We found that SbtB slowly hydrolyzes the adenine nucleotides ATP and ADP to AMP, show that this activity is redox-regulated via the R-loop, provide evidence that its deletion affects C_i acclimation and diurnal growth, and suggest that it serves to coordinate SbtA:SbtB complex formation in response to the day/night cycle.

Results

SbtB Has ATP Diphosphohydrolase (Apyrase) Activity.

In our initial structural characterization of SbtB (3), we were able to solve crystal structures of ScSbtB in apo-state as well as in complex with AMP and cAMP, which all crystallized in the same trigonal crystal form in space group P3₂, irrespective of the ligandation state. While displaying the canonical trimeric PII architecture, ScSbtB has a peculiar CGPxGC motif at its C-terminus, which is conserved in many but not all SbtB homologs in cyanobacteria (SI Appendix, Fig. S1) (3). In this motif, which we term the R-loop (for Redox-regulated loop), the two cysteines were found to form a disulfide bridge (PDBs: 5O3Q and 5ORQ). Moreover, the carboxyl terminus of the C-terminal cysteine, in some cases, is found to form a hydrogen bond with the amino terminus, which appeared to stabilize the whole R-loop assembly (3). The adenine nucleotides, either AMP or cAMP, were found to form the canonical interactions in the intersubunit clefts, with their phosphate groups mostly solvent-exposed and the T-loop largely disordered (3, see also Figs. 2 A and E and 3C). To our surprise, cocrystallization attempts with ADP (3) and also ATP, which also yielded the same trigonal crystal form, contained only AMP in all three binding sites, suggesting that the nucleotides were hydrolyzed during the time course of the crystallization experiment. To find out whether the purified recombinant Strep-tagged ScSbtB preparation contained a contaminating ATPase/ADPase activity, the ScSbtB preparation was analyzed by tandem mass spectrometry (MS/MS) (SI Appendix, Fig. S2 and Dataset S1). As no contamination with potential phosphate hydrolase activity could be detected, we speculated that ScSbtB might itself hydrolyze the adenine nucleotides, similar to what has been reported for canonical PII proteins (14). Therefore, we assayed for phosphorylase activity using ATP and ADP as substrates and confirmed slow ATP and ADP hydrolysis activity by ScSbtB, which was five fold accelerated in the presence of Mg²⁺ (Fig. 1A and SI Appendix, Fig. S3). A very similar activity was obtained using a different recombinant SbtB protein, derived from the filamentous cyanobacterium Nostoc sp. PCC 7120, indicating that the hydrolysis of adenine nucleotides is a common trait among cyanobacterial SbtB proteins. To reveal the specificity of metal ions on phosphate release, various divalent metal ions at 5 mM concentration were added to the assay (SI Appendix, Fig. S3), revealing that also Mn²⁺ and Co²⁺ stimulate phosphate release in addition to Mg²⁺, whereas all other metals were ineffective. Since Co²⁺ is toxic for cyanobacteria and cannot be found in excess inside the cell (15), we concluded that Mg²⁺ and Mn²⁺ are most likely the physiologically relevant metals used by SbtB.

Fig. 2.

Overall structure of SbtB and sequence of ATP/ADP hydrolysis with an oxidized R-loop. Top: (A) Trimeric structure of the SbtB:AMP complex. The nucleotides are bound between the subunits, and the T-loops are largely unfolded and indicated by dashed lines. The disulfide bonds in the oxidized R-loop are indicated in stick representation. Bottom: Structural snapshots (from two perspectives) along the sequence of hydrolysis when the R-loop is oxidized. (B) After the short ATP soak, fully occupied ATP is found in all three binding sites. (C) After prolonged soaking times, ATP is found to be hydrolyzed to ADP, with the β-phosphate in the same orientation as in the ATP-bound structure. (D) In the ADP soak, the β-phosphate is found in a different orientation, which might represent the orientation that needs to be assumed for ADP hydrolysis. (E) Finally, the AMP-bound structure represents the end of the full hydrolysis sequence, which can be obtained by cocrystallization with either AMP, ADP, or ATP. Gray spheres represent water molecules. Electron densities for the individual ligands are shown in SI Appendix, Fig. S4, all three chains of the short ATP soak together with sodium coordination are detailed in SI Appendix, Fig. S5.

Fig. 3.

Complex structures illustrating how the SbtA:SbtB association is regulated by nucleotide binding. (A) In the SbtA:SbtB complex, the SbtB T-loops are inserted into cavities of the individual SbtA subunits, stabilized by AMP molecules. This particular T-loop conformation is colored red and anchored by an interaction between R46 and N59 (B). In the complex structures SbtB:cAMP (C), SbtB:c-di-AMP (D), SbtB:ATP (E), SbtB:ADP (F), SbtB^defR:ATP (G), and SbtB^defR:ADP (H), this T-loop conformation is prevented in different ways. For each complex, both the detailed nucleotide binding modes and the superposition to the T-loop in SbtA-bound conformation (red) are illustrated. In the superpositions, the areas in which the SbtA-bound T-loop conformation would clash with the bound nucleotide are highlighted with a red glow. In the cAMP-bound state (C), the cAMP phosphate would preclude R46 from assuming its anchoring role, while in the c-di-AMP state (D), the base of the T-loop is found in an entirely different conformation. For the ATP and ADP state with an oxidized R-loop (short and long ATP soaks, E and F), the phosphate groups of the nucleotides would obstruct the T-loop path. The same is the case for the ATP and ADP state with a reduced R-loop (mimicked in SbtB^defR, G and H), where the T-loop is partly or completely folded around the phosphate groups: In SbtB^defR:ATP, the T-loop is fully ordered and wrapped around the phosphates, forming a sophisticated network of interactions, which is completely different from the T-loop conformation in the StbA-bound state. In analogy to the latter, R46 is also interacting with N59, but in a completely different orientation, additionally coordinating the γ-phosphate (SI Appendix, Fig. S6). In SbtB^defR:ADP, the interaction network is similar but less complete. Strikingly, R46 is in a similar orientation as in the reduced ATP state, but coordinating the β-phosphate instead of the γ-phosphate group. A superposition of all T-loop conformations is shown in SI Appendix, Fig. S8.

Fig. 1.

Apyrase activity of SbtB via phosphate release assay. (A) Phosphate release assay for ScSbtB and NsSbtB in the presence and absence of Mg²⁺ ions. The released inorganic phosphate (P_i) is shown in µM. (B) Relative phosphate release activity of different SbtB variants compared to WT ScSbtB (100%), as indicated. Values are means ± SD; n = 5 to 6 independent measurements of different SbtB purifications.

Following ATP/ADP Hydrolysis by ScSbtB in Crystallographic Snapshots.

As hydrolysis appeared to be a slow process (SI Appendix, Fig. S3), we tried to study it structurally by performing crystal-soaking experiments (13). To this aim, we incubated the P3₂ crystals of apo-ScSbtB, in which the R-loop is folded via the disulfide bridge (Fig. 2A), in their crystallization solution supplemented with either ADP or ATP for different time spans. These crystals yielded several datasets with resolutions better than 2.5 Å and well-defined electron density for the unambiguous identification of β- and γ-phosphate groups of ADP and ATP molecules (SI Appendix, Fig. S4). Together, the different structures illustrate different time points of a potential hydrolysis reaction from bound ATP over ADP to AMP: While a short (2 h) ATP soak (PDB: 7R2Y) shows clear electron density for ATP in all three binding sites (Fig. 2B), ATP is found to be hydrolyzed to ADP in a long (overnight) ATP soak (PDB: 7R2Z) (Fig. 2C), and ADP (Fig. 2D) is found mostly hydrolyzed to AMP in a (4 h) ADP soak (PDB: 7R30). We caution that the mentioned soaking times (2 h, 4 h, overnight) are not expected to correlate with the turnover times in solution but merely reflect the methodological requirements of the crystal system. Compared to the SbtB:AMP structure (Fig. 2 A and E), the complexes with ADP and ATP show additional interactions between their β- and γ-phosphate groups and the protein, which is most pronounced in the short ATP soak.

In the short ATP soak, the three chains of trimeric ScSbtB differ noticeably in the area of the effector molecule binding cleft, where the T-loop is structured to different extents. Compared to the AMP-bound structure (PDB: 5O3R), where the T-loop is disordered after residue G41, the base of the T-loop is resolved until R43 in one (Fig. 2B), and the T-loop is resolved until S48 in another chain (chain C), where it wraps around the phosphate groups (SI Appendix, Figs. S1 and S5). ATP is found in a strained conformation, in which one of the γ-phosphate oxygens (Oγ1) is forming a 2.6 Å intramolecular hydrogen bond to the 3′-hydroxyl group of the ribose moiety. The Oγ2 oxygen is forming a hydrogen bond with the G41 backbone nitrogen, and in chain C, Oγ3 and Oγ1 are forming salt bridges with the R46 guanidino group. Also, in chain C, the β-phosphate forms hydrogen bonds with the N44, V45, and R46 backbone nitrogen atoms, while the α-phosphate forms hydrogen bonds with the S42, R43, and G89 backbone nitrogen atoms and a salt bridge with the R43 guanidino group, similar to the SbtB^defR:ATP complex described later in this manuscript (Fig. 3G). In two of the chains, the coordination of ATP is completed with a metal ion bridging the β- and γ-phosphates, which is further coordinated by N44 in chain C. Based on the electron density and coordination distances of about 2.4 Å, we interpreted this metal as a sodium ion from the crystallization buffer, potentially mimicking a physiologically relevant Mg²⁺ ion. Of note, the described T-loop conformation in chain C has a weaker electron density than the rest of the protein, while the residues preceding the R-loop have a weaker electron density as well. Closer inspection reveals that T-loop residues S42 and R43 are in too-close contact to R-loop residues H102 and T103, such that both stretches cannot assume the described conformation simultaneously. Accordingly, these residues were modeled with partial occupancy, reflecting the overall weaker electron density. Taken together, we observe that the T-loop can wrap around and coordinate the phosphates of ATP, but this conformation is conflicting with the folded R-loop.

In the long ATP soak, we find that ATP was hydrolyzed to ADP in two of the chains, and the start of the T-loop is only structured until G41 in all three chains. The β-phosphates of the ADP molecules are found in the same orientation as in the ATP molecules prior to hydrolysis, but they do not form additional interactions with SbtB as the T-loop is unstructured, while the R-loop is ordered as in the other structures. In the ADP soak, however, we find ADP in another conformation (Fig. 2D). While it is already converted to AMP in two chains, in the third chain, it has the β-phosphate relocated toward the R-loop, where it interacts with two additionally folded T-loop residues, with the S42 backbone nitrogen and the R43 guanidino group. While this could possibly be an artifact of the soaking procedure, it is conceivable that this conformational switch of the ADP molecule is a necessary part in consecutive steps of the hydrolysis reactions from ATP to ADP and from ADP to AMP. In the light of these results, we concluded that SbtB might display redox-dependent ATP/ADP diphosphohydrolase (apyrase) activity. As the folded (oxidized) R-loop is conflicting with a conformation in which the T-loop is wrapping around the ATP phosphates, we assumed that its function might be to promote hydrolysis, while a reduced and thus unfolded R-loop might allow a tighter binding and thus stabilization of ATP.

T-Loop Arginines and Oxidized R-Loop Are Critical for ATP Hydrolysis.

The observed nucleotide binding modes, the stepwise hydrolysis of ATP, and the apparent incompatibility of T-loop folding with the oxidized R-loop inspired a number of experiments, for which we constructed two sets of mutants. The first set comprises point mutants in the T-loop, in which we substituted either R43, R46, or K40 by alanine, which are involved in the coordination of the β- and γ-phosphates and the metal ion coordinating these phosphates in the short ATP soak (PDB: 7R2Y), respectively. The second set was inspired by the apparent incompatibility of the T-loop with the R-loop folding. It comprises variants in which we mutated both R-loop cysteines (C105 and C110) to either alanine or serine (C105A+C110A or C105S+C110S) such that the R-loop should no longer be able to assume its folded structure due to the lack of the disulfide bond and would resemble a permanent reduced state of SbtB. In addition, a variant was created in which we truncated the R-loop from position 104 on (Δ104).

With both sets of variants, we performed phosphate release assays, and in fact, all variants showed reduced phosphate release activity, both with Mg²⁺-ATP and Mg²⁺-ADP as substrates (Fig. 1B). These results have two implications. First, they show that R43, R46, and K40 are not only important for the binding of ADP and ATP but also for their breakdown. Second, while the folded R-loop supports this breakdown, hydrolysis is efficiently inhibited when the R-loop is not folded or absent, which is especially interesting in the light of our previous hypothesis that the R-loop might function as a redox switch (3).

The T-Loop Can Adopt an ATP-Protecting Conformation in the Absence of a Folded R-Loop.

As our initial crystal structures already suggested that a structured R-loop is incompatible with the folding of the T-loop, we subsequently performed structural studies on ScSbtB variants either lacking the R-loop (Δ104) or mimicking its reduced state using the alanine or serine substitution variants (C105A+C110A or C105S+C110S). For simplicity, we will refer to both variants as SbtB^defR (for deficient in/defunctional R-loop), where applicable. With those variants, we performed cocrystallization trials with ADP and ATP, which all yielded the same crystal form in space group P4₁. For ADP, the best dataset was obtained with the Δ104 variant, which was scaled to 1.8 Å resolution (SbtB^defR:ADP), while the best dataset with ATP was obtained with the C105A+C110A variant, which was scaled to 1.5 Å resolution (SbtB^defR:ATP). The structures have one SbtB trimer in the asymmetric unit and show unambiguous electron density for their respective nucleotide in all three binding sites (SI Appendix, Fig. S4). Strikingly, the two structures are very similar to the ADP- and ATP-bound structures recently reported for SbtB from Cyanobium sp. PCC7001 (7), which belongs to a group of SbtB proteins lacking the R-loop extension (SI Appendix, Figs. S1 and S6). This might imply that SbtB does not possess apyrase activity in the absence of the R-loop. In all chains of the SbtB^defR:ATP and SbtB^defR:ADP structures, the base of the T-loop assumes a similar conformation as the one observed in one chain of the short ATP soak, which was incompatible with the folded R-loop (see above; SI Appendix, Fig. S7). However, besides numerous similarities, the ADP- and ATP-bound structures also show peculiar differences in their nucleotide binding modes and T-loop conformations (Fig. 3), as detailed in the following:

In the SbtB^defR:ATP structure (PDB: 7R31), the R-loop is largely disordered (SI Appendix, Fig. S7) and the T-loop is entirely folded in two chains, and in all three chains, all nucleotide interactions (Fig. 3G) are formed as described for the most complete chain (C) in the short ATP soak, including the sodium ion with a well-resolved complete octahedral coordination sphere (Fig. 3E and SI Appendix, Fig. S5). As this T-loop conformation wraps around the phosphate groups of the nucleotides to potentially prevent their hydrolysis, we call this the “protecting state”. Of note, the functionally important R46, which is coordinating the γ-phosphate, is additionally forming a hydrogen bond with N59, which was also observed in the short ATP soak. N59 generally forms hydrogen bonds with the hydroxyl groups of the ribose of all bound nucleotides, but in the protecting state, it has its side chain flipped to additionally bind to R46, which presumably stabilizes the T-loop in this conformation.

In the SbtB^defR:ADP structure (PDB: 7R32), the T-loop is not fully resolved, but in two chains, it is found in a conformation similar to the ATP-protecting state. ADP is found in a similar orientation as in the post-hydrolysis state, with the β-phosphate in a similar location as the ATP γ-phosphate. Additionally, R46 is found in essentially the same conformation as in the ATP-protecting state, but it is now interacting with the β-phosphate instead of the γ-phosphate, and it forms the same stabilizing hydrogen bond with N59 as in the ATP-protecting state (Fig. 3H). However, there are significant conformational differences in the T-loop residues preceding R46, which interact with both the α- and β-phosphate in the ATP-protecting state. Most prominently, R43 in the SbtB^defR:ADP complex is found in a very different location, where it is no longer interacting with the α-phosphate but with the β-phosphate (Fig. 3H). In this overall conformation, the path of the T-loop backbone, most importantly that of S42 and R43, would not necessarily clash with a folded R-loop (SI Appendix, Fig. S7). Therefore, we assume that this conformation is also possible in wild-type (WT) ScSbtB when the R-loop is folded, although it was not captured in our initial crystallization attempts.

Structural Basis for Adenine Nucleotide-Dependent SbtA:SbtB Interaction.

With our previous and current crystallographic analyses, we have delineated the binding modes of the nucleotides cAMP, AMP, ADP, ATP, and also c-di-AMP and could rationalize the influence of the R-loop on the stepwise hydrolysis of ATP to AMP. In the next step, we wanted to uncover how the binding of these nucleotides influences the complex formation between SbtB and its major target SbtA (8, 9). Therefore, we compared the different conformational states of the individual SbtB:nucleotide complexes to the recently reported structure of the SbtB:AMP:SbtA complex (Fig. 3). Within this complex, the T-loops are inserted into the individual SbtA subunits, forming an extensive interface, which is stabilized by the binding of AMP: At the base of the T-loop, the AMP phosphate group is coordinated by the S42 and R43 backbone nitrogens, promoting a sharp turn of the T-loop toward SbtA. Further, considering the structures of the different SbtB complexes, an interaction network between R43, R46, and N59 is of special interest. In the SbtB:AMP:SbtA complex, the R46 guanidino group forms similar hydrogen bonds with N59 as in the SbtB^defR:ATP and SbtB^defR:ADP structures, just that the guanidino group is in a different orientation, as the R46 backbone is not near the phosphate but buried in the interface to SbtA. On its opposite side, the R46 guanidino group forms hydrogen bonds to the R43 backbone oxygen, completing the network of interactions between these three signature residues (SI Appendix, Fig. S6). Consequently, R46 and N59 have dual roles, as they can either stabilize the ATP-protecting state or the SbtA-bound conformation, while R43 is found in a number of different interaction modes depending on the complexation state. Although the different SbtB:nucleotide binding modes revealed a number of structural differences, they have one feature in common: All but the AMP-bound state are incompatible with the T-loop conformation required to adopt the SbtB:AMP:SbtA complex (Fig. 3 A and B). For cyclic nucleotides, bound cAMP prevents the formation of the R43-R46-N59 network, as it would clash with the cAMP phosphate group (Fig. 3C), while bound c-di-AMP completely obstructs the folding of the base of the T-loop in the required conformation (Fig. 3D), independent of the redox state of the R-loop. For the binding of ADP and ATP, the mode of obstruction depends on the R-loop. When the R-loop is folded (oxidized; SI Appendix, Fig. S7), the β- and γ-phosphates collide with the necessary path of the T-loop (Fig. 3 E and F). With an unfolded (reduced) R-loop (SI Appendix, Fig. S7), the additional wrapping of the T-loop around the phosphates further prevents the formation of the SbtB:SbtA interface (Fig. 3 G and H), which is most apparent when looking at the different location and orientation of R46.

SbtB Interacts with TrxA.

As the R-loop was oxidized in all our ScSbtB structures (3, 13), we attempted crystallization trials with ScSbtB in presence of a strong reducing agent TCEP (Tris(2-carboxyethyl)phosphine hydrochloride). Furthermore, the apyrase activity was assayed after overnight dialysis in presence of 1 mM TCEP. Surprisingly, the reducing agent had no apparent effect, as we obtained the same crystals in space group P3₂ as for oxidized ScSbtB with a folded R-loop and the apyrase activity was not significantly affected (SI Appendix, Fig. S3C). We speculated that the reduction may require enzymatic catalysis by a disulfide oxidoreductase in vivo. In fact, MS/MS analysis of recombinant strep-tagged ScSbtB purified from Escherichia coli showed that the thioredoxin (TrxA) was among the most highly coenriched proteins (SI Appendix, Fig. S2 and Dataset S1). This suggested TrxA as a likely candidate for reducing the R-loop of SbtB. Indeed, Synechocystis TrxA (encoded by slr0623) is the only essential and most abundant TrxA, as it is in all oxygenic photoautotrophs (16). Similar to SbtB (13), TrxA is also responding to day/night changes of the cellular redox state (17). To examine a possible SbtB-TrxA interaction, we used our SbtB-specific bacterial adenylate cyclase two-hybrid (BACTH) interaction assay (13). Here, the T25 subunit of adenylate cyclase (Cya) was fused either N- or C-terminally to SbtB, while the T18 subunit of Cya was fused N- or C-terminally to TrxA (encoded by slr0623) (SI Appendix, Fig. S9). The established SbtB interacting T18-GlgB fusion or the leucine zipper interaction was used as positive controls (13), while an empty pUT18 vector or a T18-GlgC construct (encoding glucose-1-phosphate adenylyltransferase by slr1176) was used as negative controls. Clear interaction was observed only between the N-terminally tagged T25-SbtB and the C-terminally tagged T18-TrxA tagged on both of X-Gal or MacConkey plates (SI Appendix, Fig. S9). Furthermore, a weak interaction was observed on X-Gal plates between the C-terminally tagged T25-SbtB and the C-terminally tagged T18-TrxA, whereas no interaction was observed using the N-terminally tagged T18-TrxA under all tested conditions (SI Appendix, Fig. S9). Next, we checked if the N-terminal T25-SbtB-Δ104 and the T25-SbtB-[C105A+C110A] fusions would interact with C-terminally tagged T18-TrxA. As expected, the deletion of the R-loop abolished the interaction with TrxA, while the T25-SbtB-[C105A+C110A] fusion interacted weakly with TrxA. This result strongly supports the specificity of TrxA interacting with the R-loop of SbtB.

Furthermore, the direct physical interaction of TrxA with SbtB was analyzed by immobilizing either recombinant strep-tagged ScSbtB or His₆-tagged TrxA protein on streptavidin or Ni²⁺ magnetic beads, respectively, and incubating the immobilized-protein with the other partner, followed by successive washes to remove unbound protein. As both SbtB and TrxA have a comparable molecular weight of 12.5 to 13.5 kDa, TrxA and SbtB were identified by immunoblotting using α-poly-His and α-strep antibodies, respectively. When SbtB was immobilized on streptavidin beads, coelution of TrxA with SbtB could be detected (SI Appendix, Fig. S9) and vice versa, when His₆-TrxA was immobilized on Ni²⁺-NTA magnetic beads, WT SbtB was detected in the elution fraction of TrxA (SI Appendix, Fig. S9). Remarkably, the SbtB C105A+C110A variant showed only very weak interaction with TrxA (SI Appendix, Fig. S9), in agreement with the BACTH data, supporting our conclusion for the specificity of the TrxA-SbtB interaction to break the disulfide bond between the R-loop cysteines.

SbtB R-Loop Is Required for Proper Physiological Acclimatization of Synechocystis.

The above results clearly indicated that the SbtB R-loop might play important regulatory roles in Synechocystis physiology. To gain deeper insights, we complemented the SbtB-deficient mutant (ΔsbtB) with the site-specific SbtB variant (C105A+C110A) and with the R-loop deficient Δ104 variant (SI Appendix, Fig. S1). In Synechocystis, SbtB was shown to have a broader impact on C_i acclimation and Calvin–Benson–Bassham (CBB) cycle activation than merely regulating the SbtA activity (3, 11). But also, it plays a pivotal role in regulating diurnal metabolism via controlling glycogen synthesis (13) and promoting light activation of the CCM and CBB cycle and their inactivation in the dark (18). To find out if the R-loop plays a role in one of these processes, we first characterized the role of the SbtB R-loop in light activation of the CCM. Therefore, both SbtB variants were compared to the WT cell under either high CO₂ (HC, 2% CO₂) or low CO₂ (LC, 0.04% CO₂) and using high-light (HL, 100 µE) intensity, a condition that triggers maximum activation of CBB cycle. Both mutants grew significantly slower than the WT (Fig. 4A and SI Appendix, Fig. S10), suggesting a role for SbtB R-loop for proper C_i and light acclimation. A similar phenotype had been shown for the ΔsbtB mutant (3). However, in contrast to ΔsbtB (3), both R-loop mutants grew almost like WT cells under low-light (LL, 40 µE) and LC conditions (Fig. 4A and SI Appendix, Fig. S10). Next, we investigated the involvement of SbtB R-loop mutants in diurnal growth by exposing the cells to 12 h light/dark cycles. Again, similar to ΔsbtB mutant (13), both R-loop mutants showed a strong growth defect under day/night cycling (Fig. 4 B and C and SI Appendix, Fig. S10).

Fig. 4.

Phenotypic characterization of the ΔsbtB R-loop mutants in comparison to WT Synechocystis PCC 6803. (A and B) Specific growth rate of ΔsbtB R-loop mutants under different light and carbon supply (A) or under 3, 5, and 7 d of diurnal growth (B). (C) Growth test by drop plate assay of ΔsbtB R-loop mutants under either continuous light (Left) or a 12-h diurnal rhythm (Right). (D) Photosynthetic oxygen production and respiration of ΔsbtB R-loop mutants throughout a 12-h diurnal rhythm for 66 h. (E) Oxygen production and consumption rates were calculated based on the data from (D) for the early day (first 1 to 2 h) and the beginning of the night (first 1 h), respectively. (F) Relative glycogen levels estimated at the midday for cells growing under either continuous light or a diurnal rhythm (after 3 day/night cycles). The glycogen content was normalized to 100% of WT cells under continuous light. (F) Inset: glycogen concentration shown in µg/10⁸ cells. (G) Efficiency of PSII measured using PAM fluorometry at the midday after 7 day/night cycles. PAM fluorometry was either measured in absence (zero μE) or at constant (56 μE) actinic light. (H) Viability test for ΔsbtB R-loop mutants using drop-plate assay after 3 and 6 d of incubation in complete darkness. (I) Efficiency of PSII measured at zero µE of actinic light using PAM fluorometry during complete incubation in darkness and after light recovery. The values in this figure are means ± SD of independent biological replicates.

To gain insight into the mechanism that makes the SbtB R-loop important for diurnal growth and full activation of CBB cycle, we measured the changes in oxygen saturation in the medium during three successive day/night cycles, as proxy for photosynthetic oxygen evolution and respiration, reflecting CBB activation and inactivation, respectively. During the day, both R-loop mutants showed remarkable less oxygen production than WT cells (Fig. 4 D and E), implying that both mutants aren’t able to fully activate the CBB cycle and tuning down their oxygenic photosynthesis similar to the ΔsbtB mutant (3, 13) and explaining the reduced growth of the mutant under 100 µE (Fig. 4A and SI Appendix, Fig. S10). Upon onset of darkness, all strains started respiration, with WT cells displaying higher oxygen consumption than the mutants (Fig. 4 D and E). In cyanobacteria, the photosynthetic glycogen synthesis during the day as a carbohydrate-reserve is essential for night survival (13, 19). Since the R-loop mutants aren’t able to fully activate the CBB cycle, we assumed that glycogen synthesis could be negatively affected, which would explain the low oxygen consumption and the sensitivity of those mutants to diurnal growth. Therefore, we determined the intracellular glycogen concentration under continuous light cultivation and under diurnal growth at mid of the day in three successive day/night cycles (Fig. 4F). Compared to continuous light growth, the glycogen levels in WT cells after 3 d of diurnal cultivation were about 65% lower, consistent with the daily usage of glycogen for nighttime survival. In both SbtB R-loop mutants under continuous light, the glycogen levels were about 80% lower than in WT cells (Fig. 4F). After 3 d of diurnal growth, the levels were further reduced to only about 5% in both R-loop mutants, explaining the growth disadvantage of both mutants under diurnal cultivation (Fig. 4 B and C and SI Appendix, Fig. S10). To further elaborate on the importance of glycogen under prolonged diurnal growth, we measured photosynthetic pigments (PP) in whole cell spectra and apparent quantum yield of photosystem II (PSII) by pulse amplitude modulation (PAM) fluorometry at midday after 7 day/night cycles as an indication of viability. A clear decline in PP was observed in the mutants (SI Appendix, Fig. S10). Even more dramatically, the apparent quantum yield of PSII was strongly reduced in both mutants compared to WT cells (Fig. 4G).

Since glycogen catabolism is the major source for respiration in the dark, we characterized the phenotype of the R-loop mutants during prolonged dark incubation by drop-plate assay and measuring the PSII efficiency and PP (Fig. 4H and SI Appendix, Fig. S11). The ΔsbtB[Δ104] mutant showed a marked loss of viability already after 3 d of darkness (Fig. 4 H and I and SI Appendix, Fig. S11). However, after 6 d of darkness, both mutants weren’t able to recover and turned whitish, indicative of cell death, while the WT cells could regrow (Fig. 4 H and I and SI Appendix, Fig. S11). Remarkably, the ΔsbtB[Δ104] mutant was more sensitive than the ΔsbtB[C105A+C110A] mutant as indicated by drop-assay and recovery of PSII (Fig. 4 H and I and SI Appendix, Fig. S11).

Previously, we showed that SbtB is involved in regulating glycogen synthesis via interacting with the glycogen-branching enzyme GlgB (13). This raised the question if the R-loop might be directly involved in the interaction with GlgB. To test this hypothesis, we used our established SbtB-specific BACTH interaction assay (13). The N-terminal fusion of T25-SbtB that interacts with the N-terminally tagged T18-GlgB fusion was used as the positive control (13). In analogy, the N-terminal fusions of T25-SbtB-Δ104 and the T25-SbtB-[C105A+C110A] were tested for interaction with N-terminally tagged T18-GlgB. Surprisingly, both R-loop variants interacted with GlgB like WT SbtB (SI Appendix, Fig. S9). Together, these results indicated that the growth defect of the R-loop mutants under diurnal cycles is not mediated by a loss of GlgB interaction. But rather, it supports the broader impact of SbtB on Synechocystis physiology and points toward an unidentified SbtB R-loop-controlled process, involving either in CBB cycle or glycogen biosynthesis.

Discussion

The unexpected apyrase activity that we unveiled for ScSbtB is unusual and has a number of peculiarities and implications. While common apyrases typically hydrolyze all types of nucleotide triphosphates and diphosphates to their monophosphates, ScSbtB is highly specific for adenosine nucleotides (3, 7, 13). This preference is due to the very specific recognition of the adenine moiety in a binding pocket that is conserved in many other proteins of the PII-like family (1, 4). Intriguingly, a recent investigation of an SbtB-like protein known as carboxysome-associated PII protein (CPII) revealed that ADP could be hydrolyzed to AMP (5), although the authors could not elaborate on whether or not the hydrolysis was catalyzed by CPII. These indications suggest that SbtB proteins generally have the ability to hydrolyze adenine nucleotides to reach a thermodynamically stable SbtB:AMP state. Moreover, it was previously shown that canonical PII proteins possess very weak in vitro ATPase activity (14, 20, 21), although the physiological significance of this activity remains unclear (22). Considering the wide distribution and low sequence conservation between canonical PII and SbtB proteins, the fact that both show ATPase activity supports our previous assumption that the proteins of PII superfamily may have emerged from an ancestral nucleotide hydrolase (4). In our structural analysis, we failed to identify residues that could serve as a general base for the deprotonation of a water molecule for the nucleophilic attack on the phosphate groups, which would be conserved among SbtB proteins. The absence of a dedicated general base might explain the slow turnover of the hydrolysis reaction, which might be facilitated by the strained conformation of the bound ATP molecule, exposing the β- and γ-phosphates to hydrolytic attack. As the apyrase activity generally leads to the AMP-bound state, in which SbtB can bind and potentially occlude the SbtA channels, it seems to generally drive the SbtA–SbtB system toward the closed state.

In this context, a striking peculiarity is the redox regulation of the apyrase activity. In the oxidized state, the C-terminal R-loop is folded, with a disulfide bond formed between its two cysteines, while it is unfolded in the absence of the disulfide bond in the reduced state. When the R-loop is reduced or absent, the T-loop can assume its ATP-protecting conformation, in which the T-loop wraps around the phosphate groups of bound nucleotides, which is presumably the default in SbtB proteins that are lacking the R-loop extension. We found that the apyrase activity is significantly dampened in this state. However, when the R-loop is oxidized, the folding of the T-loop in the ATP-protecting conformation is prevented, which increases the apyrase activity significantly. Since the redox state in oxygenic phototrophs switches between the dark and light, our results suggest that ScSbtB has increased apyrase activity in the dark, when oxidized, resulting in an accelerated turnover from the ATP-bound to the AMP-bound state of SbtB, and thereby to an accelerated closure of the bicarbonate channel SbtA.

Our results imply that the R-loop is reduced enzymatically by TrxA in vivo. A hairpin structure similar to the R-loop was previously found in disulfide oxidoreductases (PDB: 3GL5) in close vicinity to the catalytic active CPxC motif, underlying the relevance of such segment for regulatory functions, presumably as a redox switch. Our results show that the major Synechocystis TrxA, is able to interact specifically with the oxidized R-loop of ScSbtB. Remarkably, TrxA also shows a light-to-dark response, similar to SbtB, such that its expression is decreased in the dark (13, 17). Thus, TrxA could be the main SbtB-reducing enzyme in the day-phase, owing to its cellular abundance. Interestingly, the trxA mutant showed a remarkable impairment of CBB cycle, underlying its importance for the photosynthetic lifestyle of cyanobacteria (16). Several CBB enzymes evolved redox-regulated C-terminal extensions formed by a C(V/I)VxVC motif, which are analogous to the CGPxGC motif of the SbtB R-loop, with the same spacing of the cysteines. Strikingly, this motif in CBB enzymes is known as a redox switch to regulate their activity via similar intramolecular disulfide bonds under successive day/night cycles (23, 24). Moreover, the disulfide bonds of CP12 proteins, small CBB-regulatory proteins (25), are also known to be redox-regulated via TrxA, and even known to form gene clusters with TrxA in cyanobacteria (26, 27). Also, a previous study showed that GlgB, one of SbtB targets (13), is a potential TrxA target (28), which further indicates the importance of TrxA in controlling the central carbon metabolism.

As the R-loop is only found in a subset of SbtB proteins, it seems to be an optional additional regulatory element that is not present in all cyanobacteria (SI Appendix, Fig. S1). Indeed, it is not involved in direct physical contacts within the SbtA:SbtB complex but only in controlling apyrase activity. Thus, the architecture of the SbtA:SbtB complex is expected to be the same in all cyanobacteria, independent of the presence or absence of the R-loop extension. Likewise, it is possibly also not a decisive element for the binding to the other SbtB targets, like the c-di-AMP-dependent interaction with the glycogen branching enzyme GlgB (13). Here, the sensing and signaling of c-di-AMP affect the glycogen synthesis, leading to reduced glycogen levels in c-di-AMP cyclase- and SbtB-deficient mutants. During the day, the cells invest energy to accumulate high intracellular bicarbonate concentrations under C_i-limiting conditions and to maintain the Na⁺-homeostasis required for HCO₃⁻ transport, to finally synthesize glycogen molecules as a reservoir for the night phases. During the day, WT SbtB is expected to reside in the reduced state without apyrase activity. Our data imply that the R-loop mutants are specifically impaired in proper activation of CBB cycle as revealed by the growth defect under saturating light. The inability of the R-loop SbtB mutants to acclimate to diurnal growth suggests that turning on apyrase activity when SbtB becomes oxidized (dark condition) is necessary for a function of SbtB, specifically during the night time, probably involving the SbtB:AMP (or SbtB:ADP) state. Therefore, it is possible that the R-loop is only used to fine-tune the interaction with a subset of unidentified SbtB targets, possibly only SbtA.

Taking into account all data that we acquired for SbtB and its interaction partners, we propose a refined model for the control of bicarbonate uptake through the SbtA–SbtB system (Fig. 5). Generally, there is always a competition between the different adenine nucleotides for SbtB binding, and an equilibrium is found on the basis of the different affinities of these nucleotides and their relative concentrations in a given situation, e.g., in the light or in darkness. However, only the SbtB:AMP complex is able to bind and occlude the SbtA channels; the binding of any other adenine nucleotide disrupts this occlusion via a modulation of the T-loop (Fig. 3). During the day, the R-loop is expected to be reduced by TrxA, shutting off apyrase activity. Thus, the occupancy of SbtB by the various adenyl nucleotides depends on their relative concentrations. While the ratio between ATP/ADP and AMP depends on the energy supply, the concentration of cAMP depends on the CO₂ supply to the cells. In vitro, SbtB has the highest affinity toward cAMP (3). We propose that during the day, most of SbtB is complexed by various adenyl nucleotides competing for the binding sites. Additionally, under high CO₂, cAMP comes into play: The major adenylyl cyclase in Synechocystis (encoded by slr1991) is activated by CO₂ but not by HCO₃⁻ (29); therefore, cAMP levels increase only in the presence of CO₂ (3) and would not be influenced by the intracellular levels of HCO₃⁻ accumulated by CCM. Another fraction of SbtB can also reside in the c-di-AMP-bound state, thereby linking C_i-acquisition with glycogen anabolism via GlgB and other potential targets. All these forms of SbtB, except for SbtB:AMP, have only weak affinity to SbtA, thereby keeping the channel in the open conformation. However, when the cells are exposed to darkness, the ATP levels are dropping, increasingly populating the SbtB:AMP state, which can bind to SbtA and close the HCO₃⁻ channels. The transition from the SbtB:ATP to the SbtB:AMP complex is additionally aided by the slow basal apyrase activity, and here, the redox regulation of the R-loop comes into play. Upon exposure to darkness, oxidation of the R-loop induces the apyrase activity, promoting transition into the SbtB:AMP state, which can tightly bind to SbtA and firmly close the HCO₃⁻ channels. The transition from the SbtB:ATP to the SbtB:AMP complex is additionally aided by dropping ATP levels. In the beginning of the day, SbtB helps to activate the CBB, in a process in which the R-loop seems critically involved. The fact that the R-loop is not found in all cyanobacteria points at different regulatory needs for controlling the light–dark transitions in the different ecological niches populated by cyanobacteria.

Fig. 5.

Scheme of nucleotide- and redox-based regulation of the SbtA:SbtB complex. See Discussion for details. Asterisks indicate that the R-loop state is irrelevant.

Materials and Methods

Full protocols are available in SI Appendix, Materials and Methods.

Generation and Purification of Recombinant Proteins.

All plasmids and primers used in this study are listed in (SI Appendix, Table S1). The recombinant C-terminal StrepII-tagged SbtB (WT or different variants) proteins from Synechocystis sp. PCC 6803 (ScSbtB) or Nostoc sp. PCC 7120 (NsSbtB) were expressed and purified as previously described (3, 30) on Strep-Tactin®/Superflow® high-capacity column (IBA), followed by size-exclusion chromatography using the ÄKTA purifier (GE Healthcare). The recombinant N-terminal His₆-tagged TrxA protein encoded by slr0623 was expressed and purified as described previously for His-tagged proteins (30–32) on the Ni²⁺-NTA column.

Phosphate Release Assay.

Phosphate concentration was determined using the ab65622 Phosphate Assay Kit (Colorimetric) [abcam®] according to the manufacturer’s instructions.

Crystallization, Crystal Handling, Data Collection, and Structure Determination.

All crystallization experiments were performed at 20 °C using the vapor diffusion method in 96-well sitting-drop plates. Apo crystals of WT ScSbtB were grown as described before (3). For cocrystallization of SbtB^defR variants with ADP (ScSbtB Δ104 variant) and ATP (ScSbtB C105A+C110A variant), 2 mM ADP or ATP were added to the protein solution and full crystallization screens were performed as described (3).

ADP and ATP soaking experiments were performed as described previously for c-di-AMP (13). The diffraction data were collected with a wavelength of 1 Å at 100 K on a PILATUS 6M-F detector at beamline X10SA of the Swiss Light Source (PSI, Villigen, Switzerland). All structures were solved based on the trigonal apo-ScSbtB structure (PDB: 5O3P) using either difference Fourier methods (for the structures in space group P3₂) or molecular replacement via MOLREP (33) (for structures in P4₁). Data collection and refinement statistics are shown in (SI Appendix, Table S2). Structural representations were prepared using PyMol.

BACTH Assay.

Plasmid construction, cell cultivation, and experimental procedure of BACTH assay were performed as described previously (13) on X-Gal and MacConkey agar plates.

Pulldowns and MS Analysis.

For pulldown experiments, 10 µM of either strep-tagged SbtB or His₆-tagged TrxA protein was incubated on either magnetic MagStrep “type3” XT beads (IBA GmbH, Göttingen) or Ni²⁺-NTA magnetic Beads (MagBeads; Genaxxon), respectively. For the detection of the protein, which coeluted with the immobilized protein, we used western blotting analysis using anti-Strep-tag II antibody (abcam) for detection of SbtB or monoclonal anti-polyHistidine-Peroxidase antibody (Sigma-Aldrich) for detection of TrxA.

To identify whether SbtB contains a protein contamination that possesses ATPase activity, recombinant purified SbtB protein from E. coli cells was-in gel digested with trypsin, then LC-MS/MS analysis was performed on a Proxeon Easy-nLC coupled to QExactive HF using 60 min gradient.

Cultivation Conditions.

All cyanobacterial growth experiments were performed in a nitrate-supplemented BG₁₁ medium. Low-carbon (LC) conditions were achieved by inoculating WT Synechocystis cells or their corresponding mutants (ΔsbtB, ΔsbtB::sbtB[C105A+C110A], and ΔsbtB::sbtB[Δ104]) in a Na₂CO₃-free medium and without any NaHCO₃ supplementation under ambient air bubbling (0.04% [v/v] CO₂). While the high-carbon (HC) conditions were achieved by bubbling the cells with 2% CO₂. Experiments under LC and HC were performed using the Multi-Cultivator MC1000 (Photon System Instruments) under either high-light (HL; 100 µE) or low-light (LL; 40 µE) conditions.

Experiments in day/night conditions were performed in a day/night chamber with a 12-h light phase (~50 μE) followed by a 12-h dark phase. To generate long dark conditions, cultures were covered from light using aluminum foil for 6 d and kept under shaking at 28 °C. The viability of the cells was checked by agar drop assays, measuring the cell pigmentation and measuring the activity of PSII using PAM fluorometry (3, 13, 15).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All data supporting the findings of this study are available within the paper. Crystallography, atomic coordinates, and structure factors have been deposited in the Protein Data Bank, www.wwpdb.org (PDB ID codes: 7R2Y, 7R2Z, 7R30, 7R31, and 7R32).