Complex Structure and Biochemical Characterization of the Staphylococcus aureus Cyclic Diadenylate Monophosphate (c-di-AMP)-binding Protein PstA, the Founding Member of a New Signal Transduction Protein Family*

Background: PstA is a cyclic di-AMP receptor and PII-like signal transduction protein. Results: Apo-PstA and complex crystal structures reveal a novel cyclic di-AMP-binding mode and induced conformational changes. Conclusion: PstA has a similar fold but distinct signal transduction properties from classic PII proteins, which function in nitrogen metabolism. Significance: Identification of common features allows for rational prediction of cyclic di-AMP-binding sites.

Bacteria have evolved intricate signal transduction systems to detect and rapidly respond to changes in the environment. These systems include two-component systems, which normally consist of a membrane-embedded histidine kinase that senses an environmental stimulus and transduces this signal to an intracellular response regulator that coordinates an appropriate transcriptional response (1). In addition, various intracellular signal transduction proteins or domains within larger proteins can directly bind intracellular metabolites and adjust their activities or that of binding partner effector proteins very rapidly without the need of de novo protein synthesis. Other molecules, which have emerged as key components in signal transduction systems, are signaling nucleotides (2). In particular, cyclic dinucleotides have gained increased attention during recent years partly due to the discovery of novel nucleotides such as cyclic diadenylate monophosphate (c-di-AMP) 4 and cyclic adenylate guanylate monophosphate (c-AMP-GMP) (3)(4)(5)(6).
c-di-AMP is a widely distributed signaling nucleotide among the Gram-positive Firmicutes and Actinobacteria (7,8). The stimuli leading to the production of c-di-AMP are currently not well understood. However, recent genome-wide protein-nucleotide interaction studies have led to the identification of several c-di-AMP-specific target proteins, providing the first clues toward pathways that are controlled by c-di-AMP. Currently known c-di-AMP-binding proteins include the cytoplasmic gating component KtrA/C of the Ktr potassium transport system, the histidine kinase KdpD, part of a two-component system that controls the expression of the Kdp potassium transport system, and CpaA, a predicted cation-proton antiporter, all identified in Staphylococcus aureus (9). Another c-di-AMPbinding protein is DarR, a transcription factor discovered in Mycobacterium smegmatis and suggested to be involved in the regulation of the expression of genes involved in fatty acid biosynthesis (10). Recently, c-di-AMP has also been described as an allosteric inhibitor of the Listeria monocytogenes pyruvate carboxylase, which is an essential metabolic enzyme that converts pyruvate to oxaloacetate (11). In the same work, four other hypothetical proteins were identified as c-di-AMP-binding proteins (11). One of these proteins of unknown function has also been previously identified as a c-di-AMP-binding protein in S. aureus and named PstA, based on a bioinformatics analysis and its annotation as a PII-like signal transduction protein (9).
The S. aureus PstA protein is the focus of this study. Using a structural, mutagenesis, and biochemical approach, we aimed to provide a better understanding on its function. Specifically, we aimed to determine whether this protein has all necessary features to serve as bona fide signal transduction proteins and how the interaction with c-di-AMP could lead to a signaling output. Very limited information is currently available on PIIlike signal proteins such as PstA; however, much is known about prototypical PII proteins. These classical PII proteins are highly conserved and are some of the most widely distributed signal transduction proteins in nature, and they play a key role in nitrogen metabolism in bacteria and archaea (12)(13)(14). PII proteins function as trimers and through protein-protein interactions control the activity of a number of different proteins ranging from enzymes, transcription factors to membrane transporters. The ability of PII proteins to interact with their effector proteins is regulated through post-translational modifications as well as the binding of ligands, including ADP, ATP, Mg 2ϩ , and 2-oxoglutarate (2-OG), where 2-OG acts as a signal of the nitrogen availability in cell (15).
Based on sequence and operon structure, classical PII proteins have been grouped into the GlnB/K, the NifI, and an uncharacterized PII-New group (16). PII-like signal transduction proteins such as PstA share too little sequence similarity with these classical PII proteins and have not been included in previous bioinformatics analyses. The best characterized PII proteins belong to the GlnB/K group, which includes the closely related GlnB and GlnK proteins. Crystal structures of several GlnB and GlnK proteins with and without bound ligand have been reported (17)(18)(19)(20)(21)(22). These studies revealed that PII proteins form homotrimers and have a highly conserved threedimensional structure. Each monomer has a compact core made up of two ␣-helices and four ␤-strands, which are connected by so called T-and B-loops (13,15). PII proteins interact through their T-loop with effector proteins, and in the absence of this protein-protein interaction, the T-loop is very flexible and is often unresolved in crystal structures (13,15). The smaller B-loop contains a Walker A motif typical for an ATPbinding site (21). Under low nitrogen conditions, when the cellular 2-OG pool is high, PII proteins cooperatively bind 2-OG, ATP, and a Mg 2ϩ ion. Two highly conserved residues, Gln-39 and Lys-58, which are located at the beginning and end of the T-loop, respectively, play a key role in ligand binding. When the 2-OG pool drops under high nitrogen conditions, PII proteins switch to the ADP-bound state with residues Gln-39 and Lys-58 now forming a salt bridge instead of interacting with the ligands (23)(24)(25). This results in the T-loop adopting two differ-ent conformations depending on the cellular 2-OG concentration, allowing PII proteins to interact or disengage with their specific effector proteins.
The binding of 2-OG is therefore a key aspect of the sensory capacity of classical PII signal transduction proteins. However, an exception has been reported for a member of the GlnB/K group, which does not appear to bind 2-OG, and hence might sense and respond to different signals (25). Also, the effector binding properties of the PII-New group are currently not known. Finally, the recent identification of the PII-like signal transduction proteins in S. aureus and L. monocytogenes as c-di-AMP-binding proteins indicates that the processes controlled by PII-like proteins and the regulatory molecules these proteins respond to are even more complex than previously anticipated.
To elucidate the signal transduction capacity of the S. aureus PstA protein as well as provide information on the mode of c-di-AMP binding and its functional consequence, we performed a structural, biochemical, and mutagenesis analysis. Our analyses revealed that PstA proteins are clearly distinct from classical and previously characterized PII proteins and provide a rationale for the c-di-AMP binding in PstA versus the ATP, Mg 2ϩ , and 2-OG binding in PII proteins. The co-complex crystal structure between c-di-AMP and the S. aureus PstA protein not only revealed the nucleotide-protein interactions but a comparison with the apo-PstA structure allowed us to suggest a mechanism for the signal transduction properties of this protein.

EXPERIMENTAL PROCEDURES
Strain and Plasmid Constructions-Bacterial strains and primers used in this study are listed in Tables 1 and 2, respectively. Escherichia coli strains were grown in Luria Bertani (LB) medium. When appropriate, cultures were supplemented with antibiotics as indicated in Table 1. The construction of plasmid pET28b-His-pstA, for the expression and purification of N-terminally His-tagged PstA (His-PstA SA ) protein from S. aureus, has been described in a previous study (9). pET28b-His-pstA was used as a template in PCRs for the construction of the plasmids used for the expression of the different His-PstA SA variants with alanine substitutions or amino acid deletions. Specifically, for construction of plasmid pET28b-His-pstA-(T28A), primer pairs ANG1591/ANG1875 and ANG1876/ ANG1592 were used in the first round of PCR. The resulting products were gel-purified and fused using primer pair ANG1591/ANG1592 in a second round of PCR. The final product was digested with enzymes NdeI and EcoRI and ligated with plasmid pET28b that has been cut with the same enzymes. Plasmids pET28b-His-pstA(GGFL-AAAA), pET28b-His-pstA(N41A), pET28b-His-pstA(T43A), pET28b-His-pstA(R67A), and pET28b-His-pstA(⌬d71-87), were constructed in a similar manner replacing primers ANG1875/ANG1876 with primers ANG1877/ ANG1878, ANG1879/ANG1880, ANG1881/ANG1882, ANG1883/ ANG1884, and ANG1886/ANG1887, respectively. For construction of plasmid pET28b-His-pstA(K2A), primer pair ANG1931/ ANG1592 was used. The resulting products were digested and ligated with cut pET28b plasmid as described above. All plasmids were initially recovered in E. coli strain XL-1 Blue, and sequences of inserts were confirmed by fluorescent automated sequencing (GATC Biotech). For protein expression and purification, the plasmids were introduced in E. coli strain BL21(DE3) ( Table 1).
Plasmid pET28b-disA BT -His was produced for the expression and purification of the C-terminally His-tagged c-di-AMP cyclase DisA from Bacillus thuringiensis (DisA BT -His) using a strategy similar to that described by Zheng et al. (26). To this end, the disA gene was amplified from B. thuringiensis BMB171 (27) chromosomal DNA by PCR using primer pair ANG1711/ 1712. The resulting product was cut with restriction enzymes NcoI and XhoI and ligated with plasmid pET28b, which has been cut with the same enzymes. The resulting plasmid, pET28b-disA BT -His, was initially obtained in E. coli strain XL-1 Blue yielding strain ANG3043. For protein expression, the plasmid was introduced into strain BL21(DE3) yielding strain ANG3048.
Protein Expression and Purification-E. coli BL21(DE3) strains (Table 1) were used for the expression and purification of the N-terminally His-tagged PstA (His-PstA SA ) protein and the different His-PstA variants. 1-liter cultures of the different strains were grown at 37°C to an A 600 of 0.5 to 0.7, and protein expression was induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside (IPTG) and cultures were incubated overnight at 16 or 18°C. Protein purifications by Ni-NTA affinity chromatography were performed as previously described with the exception that 10 mM imidazole was included in the lysis buffer (28). Proteins were further purified by size exclusion chromatography using a 50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 5% glycerol buffer system. Protein containing fractions were pooled and concentrated to around 3 mg/ml using 10-kDa cutoff filtration devices. For crystallization studies, a 20 mM Tris-HCl, pH 7.5, buffer was used for the size exclusion chromatography and the protein was concentrated to ϳ10 mg/ml. Protein concentrations were determined using the BCA Protein assay kit from Pierce. The purity of the purified proteins was assessed on Coo-

TABLE 2 Primers used in this study
Restriction sites are underlined.

Number Name Sequence
H6 , 100 mM sodium acetate pH 4.6, 30% MPD (condition 2), 200 mM Na malonate pH 4.6 and 20% PEG3350 (condition 3), 200 mM succinic acid, pH 6.8 (condition 4). The best crystals were obtained in condition 3. Data collection was performed at 100K at the SOLEIL synchrotron, beamline PROXIMA 1 from a single crystal grown in condition 3. The His-PstA SA crystals belonged to space group P321 with unit cell parameters a ϭ 56.
Data were indexed in XDS (29), scaled in SCALA (30) and R free was generated randomly in UNIQUE (31). Molecular replacement was performed in PHENIX (32) using an ensemble of each monomer of the structure of Pediococcus pentosaceus ATCC 25745 protein PEPE_1480 (PDB code 3M05), a PstA homolog, as a search model. Two datasets were collected from the same crystal and merged together to obtain high multiplicity for sulfur phasing (total rotation ϭ 360°). However anomalous signal extended only to 4.5Å hampering direct experimental phasing. Nevertheless, anomalous maps were generated in SFALLS (31) and superposed on the MR solution to guide model building starting from the sulfur-containing amino acids, followed by the building of aromatic amino acids in COOT (33). Ten cycles of rigid body refinement (10.0 -6.0Å) were followed by 10 cycles of restrained body refinement (49.18 -2.8Å) in REFMAC (34) leading to R factor ϭ 0.25 R free ϭ 0.29. TLS group analysis was performed using the TLSMDserver (35) and performed in REFMAC. Model building was reiterated in COOT and structure validation was performed in MOLPROBITY (36). Final statistics for data collection and model refinement are presented in Table 3. The structurebased sequence alignment was generated in PyMOL and analyzed with JALVIEW (37). Structural homology searches were performed using the DALI server (38) and amino acid conservation was mapped onto the crystal structure using ConSurf (39).
For co-crystallization studies, His-PstA SA at a concentration of 10 mg/ml was incubated in 20 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 buffer containing 4 mM c-di-AMP (InvivoGen) for 30 min on ice and subsequent crystallization trials were performed in condition 3, analogous to the apo-protein. Data collection was performed from a single crystal at 100K at SOLEIL synchrotron, beam line PROXIMA2. The crystals belonged to space group I121 with unit cell parameters: a ϭ 64.21Å, b ϭ 65.65Å, c ϭ 85.16Å, ␣ ϭ 90°, ␤ ϭ 95.98°, ␥ ϭ 90°. The structure was solved by molecular replacement using the apo-structure as a model in PHASER (40). Model building, refinement and validation were performed as described for the apo-structure. The final R factor and R free were 0.19 and 0.23, respectively. Superposition of the apo and complex structures was performed with SUPERPOSE (31). The structures were deposit in the protein data bank under PDB codes 4D3G (apo PstA) and 4D3H (c-di-AMP/PstA complex).
Preparation of E. coli Whole Cell Lysates-For the preparation of whole cell lysates, E. coli BL21(DE3) strains containing the empty pET28b vector, pET28b-His-pstA or derivatives thereof for the expression of the different His-PstA SA variants, were grown overnight at 30°C in LB medium. The next morning, the protein expressed was induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside, and the cultures were incubated for a further 6 h at 30°C. Bacteria from 1 ml of culture were collected by centrifugation and suspended to an A 600 of 40 in binding buffer (40 mM Tris, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 ), containing 2 mM PMSF, 20 g/ml DNase, and 0.5 mg/ml lysozyme. Cells were lysed by three freeze and thaw cycles and subsequently used in binding assays or stored at Ϫ20°C.
Synthesis of 32 P-Labeled c-di-AMP-32 P-Labeled c-di-AMP was synthesized from [␣-32 P]ATP (PerkinElmer Life Sciences) by incubating 55.5 nM [␣-32 P]ATP with 2 M purified B. thurengensis DisA BT -His protein in 40 mM Tris, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 binding buffer at 45°C overnight. The sample was subsequently incubated for 10 min at 95°C, and the Dis-A BT -His protein was removed by centrifugation at 17,000 ϫ g for 5 min; the supernatant was transferred to a new tube and stored at Ϫ20°C. The conversion of [␣-32 P]ATP to 32 P-labeled

c-di-AMP-PstA Co-complex
c-di-AMP was estimated to be 87-96%, as assessed by TLC and densitometry analysis, which was performed as described previously (9). Differential Radial Capillary Action of Ligand Assay (DRaCALA)-c-di-AMP binding to His-PstA SA or its variants was performed with DRaCALAs, the principle of which is described in Roelofs et al. (41) and was adapted by Corrigan et al. (9) to study c-di-AMP protein interactions. Briefly, E. coli whole cell lysates or 10 M solutions of purified protein (for standard assays) in binding buffer were mixed with ϳ1 nM 32 Plabeled c-di-AMP and incubated at room temperature for 5 min. Then, 2.5 l of these reactions were spotted onto nitrocellulose membranes (Hybond-ECL, GE Healthcare) and airdried, and radioactivity signals were detected and quantified as described previously (9,41). For K d measurement, 2-fold serial dilutions of purified His-PstA SA or the different variants were prepared in binding buffer (40 mM Tris, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 ) at a starting concentration of 300 M and subsequently mixed with ϳ1 nM 32 P-labeled c-di-AMP. The fractions of ligand bound and K d values were calculated as described previously (41). To determine k off values, a 10 M solution of purified His-PstA SA protein in binding buffer was mixed with 1 nM 32 P-labeled c-di-AMP and incubated at room temperature for 5 min. Then, 2.5 l of these reactions were spotted onto nitrocellulose membranes and set as time 0. Afterward, cold c-di-AMP at a final concentration of 300 M was added to the remaining sample, and after incubation for 10, 20, 30, 40, 50, 60, 90, 120, and 180 s, 2.5 l of these reaction were A 100 M solution of purified His 6 -PstA SA protein was analyzed by SEC-MALs and the molar mass and differential refractive index plotted versus the elution volume from a Superdex-200 10/300 column. The purified His 6 -PstA SA protein (calculated mass of 13.9 kDa) eluted as a 43.9-kDa complex indicating that the protein forms a trimer in solution. Of note, the first peak is likely aggregated protein, which eluted in the void volume of the column. Inset on the left shows a Coomassie-stained gel with 1 g of purified protein and molecular mass marker proteins with sizes shown in kDa.
spotted onto a nitrocellulose membrane. The fraction of ligand bound and k off values were calculated as described previously (41).
Multiangle Laser Light Scattering (SEC-MALs) Analysis-Purified His-PstA SA protein was adjusted to a concentration of 100 M in 50 mM Tris-HCl, pH 7.5, 200 mM NaCl, and 5% glycerol buffer. One hundred microliters of this solution was subjected to a size exclusion chromatography step using a Superdex 10/300 column, which was coupled to a multiangle laser light scattering detector (Wyatt Technology Corp.). The data were processed with the ASTRA 5.3.4 software and fitted according to the Zimm model for static light scattering. Experiments were repeated twice with two different protein preparations, and a representative graph is shown.
Bioinformatics Analysis-The PstA SA protein sequence (SAUSA300_0460) was used as query sequence in a PSI-BLAST search using default settings to identify homologs in the NCBI nonredundant protein sequence database. Four iterations of the search were performed at which point only the number of classical PII proteins increased. This yielded 911 sequences with a maximum expect (e) values below 3e-04 and a minimum sequence coverage and sequence identity of 60 and 30%, respectively. These sequences were exported and used in JALVIEW (37) and a multisequence alignment generated after 10 iterations and a conserved logo-sequence with Cluster Omega (42).

RESULTS
Bioinformatics Analysis of PstA Proteins-PstA, a PII-like signal transduction protein of unknown function, was identified as a c-di-AMP receptor protein in both S. aureus and L. monocytogenes (9,11). To gain insight into the phylogenetic distribution of PstA-type proteins and define common and unique features compared with classical PII proteins, a detailed bioinformatics analysis was carried out. A PSI-BLAST search was performed using the PstA protein SAUSA300_0460 from S. aureus USA300_FPR3757 (from here on referred to as PstA SA ) as query sequence. This yielded 911 sequences with maximum expect (e) values below 3e-04 and a minimum sequence coverage and sequence identity of 60 and 30%, respectively. The majority of the retrieved sequences was annotated as hypothetical proteins, although several were annotated as nitrogen regulatory PII-related proteins. Next, a multiple sequence alignment was performed with the 911 PstA-like sequences with Cluster Omega and 10 iterations, until the alignment was stable and a consensus logo motif was generated (Fig. 1A). This analysis revealed that amino acids Lys-2, Asp-12, and Arg-67 (using PstA SA protein amino acid numbers) were highly conserved and present in more than 95% of the PstA sequences (Fig. 1A). In addition, highly conserved amino acid motifs were identified between residues Thr-28 and Thr-43 with the consensus sequence of TKLXXXGGFLXXGNTT and a highly conserved c-di-AMP-PstA Co-complex 94 GGA 96 amino acid stretch toward the C-terminal end of the protein (Fig. 1A). A segment of ϳ30 amino acids between the conserved residues Arg-67 and Gly-94 showed the highest sequence variation; as discussed later, this region is likely involved in effector protein binding (Fig. 1A).
Apo Structure of PstA SA -To further characterize the PstA SA protein, we pursued the determination of its structure. The N-terminally His-tagged PstA SA protein (His-PstA SA ) was expressed and purified from E. coli strain BL21(DE3) (Fig. 1B). The His-PstA SA protein has a calculated mass of 13.9 kDa; however, SEC-MALs analysis revealed that the protein forms a 43.9 Ϯ 0.84-kDa complex in solution (Fig. 1B) suggesting that PstA SA is trimeric in solution, similar to classical PII proteins. The crystal structure of the S. aureus PstA proteins was determined at 3.0 Å resolution by molecular replacement using the structure of the Pediococcus pentosaceus ATCC 25745 protein PEPE_1480 (PDB code 3M05) as a model (Fig. 2, A and B). This protein, which is also of unknown function, shares 48% sequence identity with PstA SA . Although the PstA SA protein crystallized as a monomer in the asymmetric unit, upon inspection of the crystal packing it became clear that the protein forms trimers within the crystal lattice consistent with a trimeric state in solution (Fig. 2, C and D). The PstA SA monomer includes a core of four ␤-strands and two ␣-helices connected by two highly flexible loops (Fig. 2, A and B). The amino acids from 34 to 40, which following the nomenclature for PII proteins will be referred to as TЈ-loop residues, and residues 68 to 92, which will be referred to as BЈ-loop residues, were not visible in the electron density maps indicating that these residues are disordered ( Fig. 2A). The main inter-monomer contacts in the PstA SA trimer are provided by the interaction of ␤ 4 , from one monomer with the C terminus of the neighboring monomer, through a network of hydrogen bonds (Fig. 2C). Inspection of the surface potential of the trimer highlighted a large, accessible, and positively charged area at the monomer-monomer interface (Fig. 2D), which, similar to the recently reported structure of the L. monocytogenes pyruvate carboxylase enzyme in complex with c-di-AMP (11), could serve as a potential nucleotide-binding site.
Structural Comparison of PstA and Classical PII Proteins-Structural similarity searches using the DALI server identified PII proteins as close structural homologs to the PstA SA protein.
Structure-based alignments (Fig. 3A) and the superposition of PstA SA with members of the PII proteins family revealed a high degree of structural similarity within the core fold of the protein

c-di-AMP-PstA Co-complex
but with considerable variation in the size and sequence of the loops (Fig. 3). Specifically, the large T-loop in PII proteins, which is involved in protein-protein interactions is replaced with a shorter TЈ-loop in PstA (Fig. 3A). Furthermore, the highly conserved residues Gln-39 and Lys-58 found in PII proteins at the beginning and the end of the T-loop, respectively, are absent in PstA-like proteins (Fig. 3). These two residues play a key role in the signal transduction function of PII proteins (15,23,24). They are involved in the binding of 2-OG/Mg 2ϩ /ATP/, at a high intracellular 2-OG concentration, although at a low 2-OG concentration when the PII protein is in the ADP-bound state, these two residues form a salt bridge, leading to a drastic change in the conformation of the T-loop (Fig. 3, B and C) (15,23,24). Furthermore, the B-loop of PII proteins is replaced in PstA by a larger loop that lacks the highly conserved B-loop motif TGXXGDGKI, which is essential for ATP binding in PII proteins (Fig. 3A). In summary, the comparative sequence and structural analyses suggest that PstA SA belongs to a new family of proteins with a similar core fold but with distinct features to PII proteins. The biological function of PstA-like proteins is currently not known. However, the absence of key residues required for ATP, Mg 2ϩ , and 2-OG binding and hence the function of classical PII proteins suggest that PstA proteins have a cellular function distinct from classical PII proteins.
Interaction of PstA SA with c-di-AMP-Previous work on PstA SA has shown that the protein can bind c-di-AMP (9). This binding is specific as other nucleotides, including ATP, cyclic adenylate monophosphate (cAMP), or cyclic diguanylate monophosphate (c-di-GMP), could not compete for the binding, even when present in large excess over c-di-AMP (9). The intracellular c-di-AMP concentration in S. aureus is in the low micromolar range (9). To assess whether the interaction between c-di-AMP and PstA SA is physiological, the K d value for this interaction was determined by the differential radial capillary action of ligand assay (DRaCALA) (41) using purified His-PstA SA protein and 32 P-labeled c-di-AMP. Using this method, an interaction with a K d of 0.37 Ϯ 0.03 M was measured (Fig.  4A). This interaction is in a physiologically relevant range and comparable with previously reported K d values for other c-di-AMP target proteins of S. aureus and L. monocytogenes, which were in the range of 0.1 to 8 M (9 -11). The k off value for c-di-AMP and His-PstA SA in the presence of 300 M cold c-di-AMP was determined as 0.076 Ϯ 0.005 s Ϫ1 (Fig. 4B), which corresponds to a half-life of 9.1 s, indicating a rapid equilibrium of bound and free protein.
PstA SA Structure in Complex with c-di-AMP-To gain molecular insights into the nucleotide protein interactions, the c-di-AMP⅐PstA SA complex structure was pursued. To this end, the His-PstA SA protein was incubated with 4 mM c-di-AMP and subsequently co-crystallized. The protein crystallized as a trimer in the asymmetric unit, and the crystal structure of the protein⅐ligand complex was solved at 2.0Å resolution by molecular replacement using the apo structure as a model (Fig. 5A). The c-di-AMP bound at the monomer-monomer interface with three molecules bound per trimer (Fig. 5B). Superposition of the ligand⅐complex structure with the apo structure revealed no major changes within the core of the protein and a root mean square deviation (r.m.s.d.) for the C␣ atoms of 0.9 Å for the trimer. The main differences between the unbound and bound structures was in the TЈ-loop, which was disordered in the apo structure but became structured in the c-di-AMP⅐PstA SA complex, thereby providing many stabilizing interactions with the ligand, particularly through Phe-36 (see detailed description below). In addition, the large BЈ-loop (amino acids 68 -92), which was unresolved in the apo structure, was more ordered in the complex, forming two anti-parallel ␤-strands (Fig. 5A). In the c-di-AMP⅐PstA SA complex structure, a salt bridge is formed between the highly conserved amino acids Asp-12 in ␣-helix 1 and Arg-67 at the beginning of the BЈ-loop (Fig. 5, A and C). In the apo structure no salt bridge is observed, as only density for the C ␤ is present for Arg-67 (Fig.  5C).
Clear density for the c-di-AMP ligand was visible at each monomer-monomer interface within the PstA SA trimer (Fig.  5D).Thec-di-AMP-bindingmodeatthethreemonomer-monomer interfaces is equivalent, and therefore only the interaction at the monomer A-monomer B interface will be discussed. The c-di-AMP-binding site is overall positively charged due to residues Asn-24, Asn-41, and Gln-108, with the latter two residues being in hydrogen bonding distance with the ligand (Fig. 5E). Both phosphate groups of the ligand form oxy-anion holes; one such anion-stabilizing hydrogen-bonding pocket is formed by the interaction of the PO 4 group 1 with two water molecules and the backbone amides of Phe-36 and Leu-37 (monomer B), DRaCALAs were performed with 32 P-labeled c-di-AMP and purified His-PstA SA at a starting protein concentration of 300 M, and fraction-bound and K d values were determined from the curve as described previously (41). The average values from three independent experiments are plotted with standard deviations. B, dissociation curve and k off determination for c-di-AMP and purified His-PstA SA . A 10 M solution of purified His-PstA was incubated with 32 P-labeled c-di-AMP. At time 0, 300 M cold c-di-AMP was added, and the displacement of 32 P-labeled c-di-AMP was measured at 10-, 20-, 30-, 40-, 50-, 60-, 90-, 120-, and 180-s time points. Fraction-bound values were plotted, and the k off value was determined as described previously (41). The average values from three independent experiments are plotted with standard deviations.

c-di-AMP-PstA Co-complex
whereas another is formed by the PO 4 group 2 through hydrogen bonding with Gln-108 and two water molecules (Fig. 6). The adenine ring A␣ is held in place by a face-to-face stacking interaction with Phe-36 (monomer B) and a -cation interaction with Arg-26 (monomer A), whereas the adenine ring A␤ forms a face-to-edge stacking interaction with Phe-99, and three hydrogen bonds with the amide and carbonyl group of Gly-47 (monomer A) and the side chain of Thr-28 (monomer A). Asn-41 forms a hydrogen bond with the ribose ring, and additional hydrogen bonds are formed between surrounding waters, which are conserved across three monomers, and the oxygen and nitrogen groups of the ribose and adenine rings, respectively (Fig. 6). It is notable that although each monomer provides approximately half a ligand-binding site, the interactions with the ligand are very different in each half-binding site.
Mutagenesis Analysis Confirms Essential Function of Conserved c-di-AMP-interacting Residues-The sequence analysis (Fig. 1A) and the c-di-AM⅐PstA SA complex structure (Figs. 5 and 6) highlighted several conserved residues involved in c-di-AMP binding. To confirm the physiological relevance of the c-di-AMP-binding site observed in the complex structure, a mutagenesis analysis was performed. Specifically, the con-served amino acids Thr-28 and Asn-41 mutated to alanines as well as the GGFL amino acids (residues 34 -37) that form the small TЈ-loop of PstA proteins were simultaneously mutated to alanines. Variants with these mutations were expected to be compromised in c-di-AMP binding as these residues make direct contact with the ligand (Fig. 6). As control, other highly conserved residues Lys-2 and Thr-43 and residues Asp-12 and Arg-67 (forming a salt bridge upon ligand binding but not in direct contact with the ligand) were mutated to alanines (Figs. 5A and 7A). Finally, a variant with a deletion of the large BЈ-loop, which distinguishes PstA proteins from classical PII proteins but is not expected to be involved in c-di-AMP binding, was constructed (variant ⌬71-87 with deletion of amino acids 71-87) (Fig. 7A). The His-PstA SA protein and the different His-PstA SA variants K2A, D12A, T28A, N41A, T43A, GGFL-AAAA, R67A, and ⌬71-87 were overexpressed in E. coli; whole cell lysates were prepared, and c-di-AMP binding was assessed by DRaCALAs as described previously (Fig. 7, B and C) (9). Reduced binding was seen for His-PstA SA variants T28A, N41A, and GGFL-AAAA, although the binding was not markedly affected for variants K2A, D12A, T43A, R67A, and ⌬71-87 (Fig. 7, B and C). To confirm this, the different variants

c-di-AMP-PstA Co-complex
were purified and used in DRaCALAs (Fig. 8). The His 6 -PstA SA ⌬71-87 protein precipitated during our standard purification process and could not be further analyzed. However, the protein showed normal binding to c-di-AMP using whole cell lysates, indicating that the large BЈ-loop was not involved in c-di-AMP binding. Identical c-di-AMP binding results were obtained for the purified His-PstA SA variants as with whole cell lysates, with variants T28A, N41A, and GGFL-AAAA showing a binding defect (Fig. 8B). Next, binding affinities were determined with the purified proteins. Initial binding studies showed that c-di-AMP bound to variants His-PstA SA K2A, D12A, and R67A with K d values of 0.34 Ϯ 0.04, 0.32 Ϯ 0.05, and 0.48 Ϯ 0.04 M, which are similar to that of wild type His 6 -PstA SA . A 5-fold increase in the K d value was observed for the T43A variant, to which c-di-AMP bound with a K d value of 1.8 Ϯ 0.2 M as compared with wild type His-PstA SA with a K d value of 0.37 Ϯ 0.03 M; however, the binding still reached saturation at a protein concentration above 20 M (Fig. 8C). The protein variants with the least binding capacity were the N41A variant with a K d of 9.6 Ϯ 2.3 M and the T28A and GGFL-AAAA variants, for which no K d values could be determined as the binding did not reach saturation even at a protein concentration of 300 M. Therefore, in agreement with interactions seen in the co-complex structure, conserved residues, Thr-28, Asn-41, and the TЈ-loop residues GGFL are important for c-di-AMP binding. Although other highly conserved residues, including residues Asp-12 and Arg-67, which form a salt bridge in the ligandbound structure, are dispensable for c-di-AMP binding. These data suggest that the physiologically relevant c-di-AMP-binding site was revealed in the complex structure.

DISCUSSION
Several c-di-AMP receptor proteins have now been identified (9 -11, 43); however, so far only the structure of the L. monocytogenes pyruvate carboxylase (PC) has been solved in complex with c-di-AMP (11). Here, we present the structure of the second target protein, namely the S. aureus PII-like signal transduction protein PstA, in complex with c-di-AMP (Figs. 5 and 6). As discussed below, although the residues involved in nucleotide binding are very distinct between these two proteins (Fig. 9) and a classical consensus sequence cannot be established, common features for c-di-AMP-binding sites can now be proposed. In addition, a comparative analysis of the apo and complex structures allows us to present a model on how this protein functions as a signal transduction protein.
Although PstA is annotated as PII-like signal transduction protein, our structural analysis together with a detailed bioinformatics analysis suggest that PstA SA is a prototype of a new protein family with clearly distinct features from classical PII proteins. This new family of proteins currently contains around 900 members in the nonredundant protein database (Figs. 1A and 3A). PstA homologs are present mainly in the Firmicutes, with some homologs present in green non-sulfur bacteria, Mycoplasmas, Spirochetes, and other unknown/unclassified bacteria. This is consistent with the identification of predicted c-di-AMP cyclases among most bacteria belonging to the Firmicutes group. However, not all bacteria that can produce c-di-AMP possess PstA homologs. Notably, a PstA homolog is present in Bacillus subtilis 168 but absent in the pathogenic strains Bacillus cereus, Bacillus anthracis, and B. thuringiensis. Conversely, the thermophilic bacteria Sphaerobacter thermophilus DSM 20745 and Thermomicrobium roseum DSM 5159 encode three PstA orthologs, whereas the majority of other bacteria contained only one PstA-type protein.
Comparison of the crystal structures of PII proteins and PstA revealed that they share the same core fold. However, as shown here, the two main loops, named T-and B-loops in PII proteins and TЈ-and BЈ-loops in PstA, are reversed in length and as we would speculate also in function. In PII proteins, the B-loop is short and contains the conserved Walker A motif sequence FIGURE 6. c-di-AMP/PstA SA hydrogen bonding network. A, c-di-AMP-binding site with amino acid residues from monomer A shown in green and monomer B shown in yellow. Residue Phe-36 forms a face-to-face -stacking with adenine, and the other residues Leu-37, Asn-41 (monomer B), Thr-28, Gly-47, and Gln-108 (monomer A) form hydrogen bonds with different parts of the c-di-AMP ligand. Additional residues shown in the structure are Phe-99 (monomer B) and Arg-26 (monomer A), forming an edge-to-face -stacking and a cation-interaction, respectively. Hydrogen bond interactions are indicated with dotted lines and interactions with double-head arrows. B, schematic representation of the hydrogen bond interactions between c-di-AMP-binding site residues. Residues from monomer A are shown in green and from monomer B in yellow. The water molecules (indicated as red W) forming hydrogen bonds are also shown in this schematic diagram. The details of these interactions are described under "Results," and adenines are labeled A␣ and A␤, and PO 4 groups 1 and 2, and will be referred to as such in the text. TGXXGDGKI required for ADP or ATP binding, whereas in PstA the short TЈ-loop contains the GGFL residues, which are part of the highly conserved TKLXXXGGFLXXGNTT motif, which is critical for c-di-AMP binding. The large T-loop in PII proteins is involved in protein-protein interaction and changes its conformation upon ligand binding. Two highly conserved PII protein residues at the base of the large T-loop, Gln-39 and Lys-58, interact with 2-OG when it is at a high cellular concentration, and therefore bound to the protein, or form a salt bridge when the cellular concentration is low and the protein is in an unbound state. This results in a conformational change in the T-loop allowing it to interact or disengage from its effector protein, thus providing a structure-based mechanism for the signal transduction properties of PII proteins in response to 2-OG levels. By obtaining the PstA apo-and c-di-AMP complex structures, we can now propose a structure-based model of how PstA functions as a signal transduction protein. In the apo-PstA structure both the TЈ-and BЈ-loops were unstructured, indicating that they are highly flexible in the nucleotide-free state (Fig. 3). Upon c-di-AMP binding, the TЈ-loop becomes stabilized and structured through interactions with the nucleotide. The nucleotide-induced ordering of the TЈ-loop results in small structural perturbations throughout the PstA trimer, and in the complex structure a salt bridge between Asp-12 at the end of ␣-helix 1 and Arg-67 at the beginning of the BЈ-loop was observed (Fig. 5). In the apo structure, the side chain of Arg-67 is not completely visible in the electron density maps suggesting that it is highly flexible (Fig. 5C). We speculate that the formation of the Asp-12-Arg-67 salt bridge could be associated with the stabilization of the BЈ-loop, which forms two anti-parallel ␤-strands that protrude from each monomer in the trimer (Fig. 5B). However, further work is needed in the future to experimentally test this hypothesis. As shown in this study, neither residue Asp-12 nor Arg-67 is essential for c-di-AMP binding (Figs. 7 and 8); however, both residues are highly conserved within the PstA proteins indicating a functional importance (Fig. 1). By analogy to PII proteins where residues Gln-39 and Lys-58 stabilizes the conformation of the T-loop, it can be speculated that residues Asp-12 and Arg-67 might play a similar role in stabilizing the BЈ-loop in PstA proteins. Again, similar to PII proteins, the conformational change in the large BЈ-loop would presumably provide an interaction surface for downstream effector proteins and in this manner affect their activity. In the absence of c-di-AMP, such interactions would be disengaged or prevented, thus establishing an elegant nucleotide-dependent signaling cascade. However, currently neither the cellular function of the S. aureus PstA protein nor any other PstA homologs are known nor have any interaction partner of PstA proteins been identified. It is notable that the BЈ-loop displays large variations among PstA homologs (Fig. 1A), and this might allow different members of this protein family to interact with diverse classes of effector proteins, similar to classical PII proteins.
The apo-PstA structure was solved using the deposited P. pentosaceus ATCC 25745 protein PEPE_1480 structure as a model, which belongs to the same protein family. In the Pediococcus structure, the TЈ-loop and a large part of BЈ-loop are structured, which we would now refer to as being in the "ligandbound state." This state was observed even though the protein crystals were grown in the absence of c-di-AMP. However, the The only other structure of a c-di-AMP bound protein complex is that of the L. monocytogenes PC enzyme. This enzyme is a tetramer composed of two layers of dimers with two nucleotide-binding sites (11). The c-di-AMP-binding sites are at symmetrical monomer-monomer interfaces, with the same set of amino acids from each monomer providing a half-binding site, consistent with the 2-fold symmetric c-di-AMP nucleotide ( Fig. 9) (11). Specifically, a tyrosine residue from each monomer provides -stacking interaction with the adenine rings, and other tyrosine and serine residues form hydrogen bonds with the terminal oxygen atoms of the phosphate groups ( Fig. 9) (11). The binding mode observed in the Listeria enzyme is similar to the previously reported binding of c-di-GMP with the eukaryotic immune receptor protein STING (44 -46). However, no clear specificity determinants can be observed for preferential interaction of the PC enzyme of L. monocytogenes with c-di-AMP versus other cyclic dinucleotides such as c-di-GMP. As the PstA protein is trimeric (Figs. 2 and 5), this results in an inherent asymmetry of the binding site with different amino acids from each monomer contributing to ligand binding (Figs. 6 and 9). The bases of nucleotide ligands often form face-to-face -stacking interactions with aromatic amino acids in the receptor. In the c-di-AMP⅐PstA complex, however, only one of the adenine bases (A␣) forms a face-to-face -stacking with Phe-36, located within the highly conserved TЈ-loop, whereas the second adenine (A␤) forms a face-to-edge stacking interaction with Phe-99; both of these amino acids are on the same monomer. In addition to the face-to-edge stacking interaction, the second adenine base (A␤) forms additional hydrogen bond interactions, including between its amino group and the amide and carbonyl groups of Gly-47 (Figs. 6 and 9). The amino group in the adenine base would be replaced with a carbonyl group in a guanine base, and therefore such an interaction would be energetically unfavorable. In addition, the amino group in a guanine base would clash with the amide backbone of Leu-35, thus providing a rationale for the specific binding of c-di-AMP to PstA.
By comparing the PstA and PC ligand-bound structures, it is apparent that c-di-AMP-binding sites are not conserved at the amino acid sequence level; thus, establishing a universal sequence motif for predicting c-di-AMP-binding sites in other proteins may not be possible (Fig. 9). Despite this, a number of conserved features of c-di-AMP-binding sites can now be identified, including a surface-accessible binding pocket formed between two protomers, with an overall positive charge (as observed for both PC and PstA), aromatic amino acids, such as tyrosine or phenylalanine (at least one in asymmetric binding sites and likely two in symmetric binding site), and additional conserved amino acids, such as threonine, serine, asparagine, or glutamine, which can provide for hydrogen bond interactions with the phosphate and ribose groups. Therefore, our study not only provides insight into how members of the PstA protein family might function as signal transduction proteins, but also highlights conserved features of c-di-AMP-binding sites that will aid future studies and A, Coomassie-stained gel with purified His-PstA SA protein (WT) and different protein variants containing the amino acid substitutions as indicated above each lane. B, c-di-AMP binding assays. DRaCALAs were performed with 32 Plabeled c-di-AMP and purified His-PstA SA protein (WT) or variants with the indicated amino acid substitution. The proteins were used at a concentration of 10 M, and as negative control (NC) no protein was added to the binding assays. Representative DRaCALA spots are shown, and the fraction bound was calculated as described previously, and the average values and standard deviation of two independent experiments were plotted. C, binding curve and K d determination for c-di-AMP and purified His-PstA SA or protein variants with the indicated amino acid substitutions. K d values were determined from the curve as described previously (41). c-di-AMP-PstA Co-complex allow for a more rational approach for identifying c-di-AMPbinding sites within target proteins.