Molecular basis for the recognition of cyclic-di-AMP by PstA, a PII-like signal transduction protein

Cyclic-di-AMP (c-di-AMP) is a broadly conserved bacterial second messenger that is of importance in bacterial physiology. The molecular receptors mediating the cellular responses to the c-di-AMP signal are just beginning to be discovered. PstA is a previously uncharacterized PII-like protein which has been identified as a c-di-AMP receptor. PstA is widely distributed and conserved among Gram-positive bacteria in the phylum Firmicutes. Here, we report the biochemical, structural, and functional characterization of PstA from Listeria monocytogenes. We have determined the crystal structures of PstA in the c-di-AMP-bound and apo forms at 1.6 and 2.9 Å resolution, respectively, which provide the molecular basis for its specific recognition of c-di-AMP. PstA forms a homotrimer structure that has overall similarity to the PII protein family which binds ATP. However, PstA is markedly different from PII proteins in the loop regions, and these structural differences mediate the specific recognition of their respective nucleotide ligand. The residues composing the c-di-AMP binding pocket are conserved, suggesting that c-di-AMP recognition by PstA is of functional importance. Disruption of pstA in L. monocytogenes affected c-di-AMP-mediated alterations in bacterial growth and lysis. Overall, we have defined the PstA family as a conserved and specific c-di-AMP receptor in bacteria.


Introduction
Bacteria utilize a wide array of nucleotide-based signaling molecules. Cyclic-di-AMP (c-di-AMP) is a broadly conserved and functionally diverse cyclic-dinucleotide second messenger that is found in many bacterial species. The levels of c-di-AMP in the cell are controlled through the opposing activities of diadenylate cyclases (DACs) which synthesize c-di-AMP from two molecules of ATP or ADP (Witte et al. 2008;Bai et al. 2012) and DHH/ DHHA1 or HD domain containing phosphodiesterases which degrade c-di-AMP to pApA or AMP (Rao et al. 2010;Bai et al. 2013;Huynh et al. 2015). Cyclic-di-AMP is of fundamental importance for microbial physiology including bacterial growth and metabolism, cell morphology, potassium homeostasis, stress responses, antibiotic resistance, and virulence (Corrigan et al. 2011;Oppenheimer-Shaanan et al. 2011;Luo and Helmann 2012;Pozzi et al. 2012;Smith et al. 2012;Mehne et al. 2013;Campos et al. 2014;Sureka et al. 2014).
Several macromolecules such as the metabolic enzyme pyruvate carboxylase (Sureka et al. 2014), the ydaO riboswitch family (Nelson et al. 2013), the transcription factor DarR (Zhang et al. 2013), and proteins involved in potassium uptake Bai et al. 2014) have been identified as c-di-AMP receptors in bacteria. In addition, the innate immune system in humans can detect secreted c-di-AMP during bacterial infection through the cytosolic surveillance protein STING (stimulator of interferon genes), leading to a host type I interferon response (Woodward et al. 2010;Burdette and Vance 2013). Recent studies on c-di-AMP signaling have begun to unravel some of the molecular mechanisms through which c-di-AMP exerts its physiological effects in bacteria. Crystal structures of c-di-AMP bound to the metabolic enzyme pyruvate carboxylase (Sureka et al. 2014) and the ydaO riboswitch (Gao and Serganov 2014;Jones and Ferre-D'Amare 2014;Ren and Patel 2014) have provided insights into the mechanism by which c-di-AMP is recognized by its receptors.
PstA (P II -like signal transduction protein A) is a previously uncharacterized protein that has been reported to bind c-di-AMP in Staphylococcus aureus  and Listeria monocytogenes (Sureka et al. 2014), and its sequence is highly conserved among the collection of bacterial species (Fig. 1A). PstA contains a DUF970 domain which is predicted to be structurally homologous to the nitrogen regulatory P II proteins ). P II proteins are able to sense ATP, ADP, and 2-oxoglutarate levels in the cell Huergo et al. 2013), and these various ligands modify the P II structure promoting its binding and regulation of protein targets such as the ammonium channel AmtB (Conroy et al. 2007;Gruswitz et al. 2007). There have been many reported crystal structures of P II proteins bound to its ligands, and the binding pocket for ATP has been particularly well characterized (Xu et al. 1998;Sakai et al. 2005;Shetty et al. 2010). In contrast, PstA was found to bind specifically to c-di-AMP but not ATP (Sureka et al. 2014), although the structural basis for this selectivity was not known. Here, we report the crystal structures of PstA from L. monocytogenes (LmPstA) bound to c-di-AMP and in the apo form. A comparison of these structures to the ATPbound P II proteins reveals significant differences in the architecture of the binding pocket for each ligand, providing a structural basis for the specific binding of c-di-AMP to PstA. We also find that the c-di-AMP binding pocket in PstA is highly conserved (Fig. 1A), suggesting that this interaction is of importance in bacterial physiology.

Experimental Procedures
Protein expression and purification L. monocytogenes pstA (lmo2692) was subcloned into a pET20b vector with a C-terminal His-tag. This construct was transformed into BL21(DE3) Rosetta cells. The cells were cultured in LB (Luria-Bertani) medium with 35 mg/ L kanamycin and 35 mg/L chloramphenicol and were induced for 14 h with 1 mmol/L Isopropyl b-D-1-thiogalactopyranoside at 20°C. The protein was purified through nickel-agarose affinity chromatography followed by gel filtration chromatography (S-300, GE Healthcare, Piscataway, New Jersey, USA). The purified protein was concentrated to 20 mg/mL in a buffer containing 20 mmol/L Tris (pH 8.0), 150 mmol/L NaCl, 5% (v/v) glycerol, and 5 mmol/L dithiothreitol, flash-frozen in liquid nitrogen and stored at À80°C. The C-terminal hexa-histidine tag was not removed for crystallization.

Crystallization
LmPstA crystals were grown by the sitting-drop vapor diffusion method at 20°C. For the c-di-AMP complex, the protein at 20 mg/mL was incubated with 2.5 mmol/L cdi-AMP for 30 min at 4°C before setup. The reservoir solution contained 12% (w/v) PEG3350 and 0.1 mol/L sodium malonate (pH 4.0). The crystals appeared within 1-2 weeks. The crystals were cryoprotected in the reservoir solution supplemented with 15% (v/v) glycerol and were flash-frozen in liquid nitrogen for data collection at 100 K. For the apo LmPstA structure, the protein at 10 mg/mL was mixed with a reservoir solution contained 20% (w/v) PEG3350, 0.1 mol/L bis-Tris (pH 5.5), and 0.1 mol/L ammonium acetate. The crystals appeared within 1 week and grew to full size within a few more days. They were cryoprotected in the reservoir solution supplemented with 10% (v/v) glycerol and flash-frozen in liquid nitrogen.

Data collection and structure determination
All X-ray diffraction data were collected at the X29 beamline at the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory. The diffraction images were processed using HKL3000 (Minor et al. 2006). The structures were solved using the molecular replacement method with the program Phaser (McCoy et al. 2008), using a PstA homolog from Pediococcus pentosaceus (PDB code 3M05) as the search model. Manual rebuilding was carried out in Coot (Emsley and Cowtan 2004) and refinement was done with the program Refmac (Murshudov et al. 1997). Coordinates and structure factors have been deposited in the Protein Data Bank with accession codes 4RWW and 4RWX.

Site-directed mutagenesis
Mutants were made using the QuikChange kit (Agilent Technologies, Santa Clara, California, United States) and confirmed by sequencing. The primers used for generation of amino acid variants are listed in Table 1. PstA containing point mutants were expressed and purified using the same protocol as the wild-type (WT) protein.

Construction of ΔpstA strain
Deletion of the chromosomal copy of pstA was accomplished using the pKSV7-oriT plasmid. Briefly, 1000 base pairs flanking the 5 0 and 3 0 ends of the pstA gene were amplified and subsequently combined by splicing by overlap extension (SOE) PCR using the primers described in Table 1. A total of six amino acids from the original open reading frame were retained to limit disruption of downstream genes. The SOE amplified product was digested with SacI and PstI and ligated into similarly digested pKSV7-oriT. The plasmid was then conjugated into L. monocytogenes WT or cΔdacA strains through the donor strain E. coli SM10. Subsequent selection for integration, plasmid excision, and plasmid curing were as described previously (Camilli et al. 1993).

Complementation of ΔpstA strain
For complementation, pstA was expressed under the constitutive hyper-Pspac promoter. The pstA gene was amplified using the primers GTG GTG GTA CCT TTT AGG AAG GGT TGA TTT CTT TG and GAT TGC TGC AGC TAG TGA TGA TGA TGA TGA TGA AAA TGA TGG AAA CTC TCA A and the PCR product was digested with KpnI and PstI restriction enzymes and ligated into the plasmid pJW361 (Sureka et al. 2014) digested with the same enzymes. The resulting plasmid was then electroporated into different ΔpstA strains.

Tissue culture and infection assays
Macrophage growth curves, DNA release, and plaque assays were performed in J2 immortalized marrowderived macrophages (BMMs) (Sauer et al. 2010) or L2 mouse fibroblasts as indicated and described previously (Sun et al. 1990).

Bacteriolysis and antibiotic susceptibility
Bacteriolysis as measured by b-galactosidase release and sensitivity to cefuroxime determined by the zone of inhibition in a disk diffusion assay were performed as described previously (Witte et al. 2013).

Biochemical and genetic characterization of PstA
We previously identified LmPstA as a c-di-AMP binding protein utilizing a chemical proteomics approach (Sureka et al. 2014). It binds c-di-AMP specifically with a K d of 1.4 lmol/L, determined with DRaCALA (differential radial capillary action of ligand assay), which is consistent with the levels of c-di-AMP in bacterial cells . We analyzed the sequence database to assess the conservation of PstA in bacteria, and found that it is widely distributed in Firmicutes (low GC content Grampositive bacteria), but is completely absent in Actinobacteria (high GC content Gram-positive bacteria) and Gram-negative bacterial species. The sequence of PstA consists of 109 residues and is generally well conserved between species (Fig. 1A). LmPstA has 16% and 20% sequence identity with the Escherichia coli P II proteins GlnK and GlnB, respectively.
We also analyzed the genomic context of the pstA gene in the bacterial species L. monocytogenes, S. aureus, Bacillus subtilis, Enterococcus faecalis, Clostridium botulinum, and Lactobacillus brevis. In all of these species, pstA is invariably found in an operon with the gene encoding thymidylate kinase, an enzyme involved in the purine salvage pathway. In several of these bacteria, a gene encoding an ornithine-arginine-lysine decarboxylase is also part of this operon. Overall structure of PstA and comparison with P II proteins We cocrystallized LmPstA with c-di-AMP and determined the structure of the complex at 1.6 A resolution by the molecular replacement method (Fig. 1B). A homolog of PstA from P. pentosaceus (PDB code 3M05), which shares 56% sequence identity with LmPstA, was used as the search model. We also determined the apo structure of LmPstA at 2.9 A resolution. There is a trimer of LmPstA in the asymmetric unit for both crystals, and LmPstA migrated as a trimer on a gel filtration column. The final atomic models have good agreement with the X-ray diffraction data and the expected geometric parameters ( Table 2). All of the residues in both structures are in the favored region of the Ramachandran plot. The structure of each LmPstA monomer is composed of a four-stranded antiparallel b-sheet packed against two a-helices on one face ( Fig. 2A). A fifth b-strand from a neighboring monomer of the trimer joins the b-sheet. We compared LmPstA to the crystal structure of a GlnK P II protein bound to ATP from Mycobacterium tuberculosis (Mtb P II , PDB code 3LF0), which is structurally representative of the larger P II family of proteins. The architecture of the LmPstA monomer is largely similar to the Mtb P II monomer (Shetty et al. 2010) with a r.m.s. (root mean square) distance of 1.0 A for their Ca atoms when the monomers are overlaid without the loop regions ( Fig. 2B and C). However, several of the loops that connect the various b-strands and a-helices have notable differences. For consistency, the naming convention for the P II pro-tein loops will be used for the description of PstA. In Mtb P II , the T-loop (residues 36-55) is a 20-residue flexible extension located in the loop connecting strands b2 and b3. The B-loop (residues 82-87) is a smaller loop connecting a2 and b4 (Fig. 2B). There is also a C-loop at the C-terminus of the protein. In PstA, however, the Bloop (residues 65-94) has a 30-residue extension while the T-loop (residues 35-38) is much shorter and LmPstA does not contain a C-loop (Figs. 1A and 2A). The B-loop forms a twisted two-stranded b-sheet (residues 67-71 and 89-93) connected by a long looping turn (residues 72-88). The two-stranded b-sheet structure of the B-loop of LmPstA is similar to the T-loop of Mtb P II . However, the length of the turn connecting the two b-strands is much shorter in the Mtb P II T-loop.
The three LmPstA monomers are almost identical in structure with an r.m.s. distance of 0.3 A for their Ca atoms, with the exception of the B-loop which shows variations among the three monomers, reflecting its flexibility ( Fig. 2A). Similarly, crystal structures of P II proteins also show T-loops adopting various conformations (Sakai et al. 2005). However, the overall orientation of the LmPstA B-loop in relation to the core trimer structure is comparable between the three monomers (Fig. 1B). The B-loop has several intermolecular interactions with the B-loops from neighboring trimers, so crystal contacts likely contribute to some degree to the observed conformations of this loop. A part of the B-loop in two of the monomers is disordered. In the apo LmPstA structure, all of the T-loops and B-loops have disordered regions. In the LmPstA trimer, the b-sheet of the three monomers form the core, and the a-helices are on the periphery (Fig. 1B). The overall organization of this trimer is similar to that of Mtb P II and the r.m.s. distance is 0.9 A for their Ca atoms when the trimers are overlaid without the loop regions.

The c-di-AMP binding mode in PstA
Three c-di-AMP molecules are bound to symmetrically equivalent locations on the LmPstA homotrimer (Fig. 1B), located in the inter-subunit cleft between b1 and b4, and the T/B-loops of one monomer and a1, b2-b3, and b5 of a neighboring monomer (Fig. 3A). A fourth c-di-AMP molecule mediates contacts between LmPstA trimers in the crystal, which is likely a crystalpacking artifact and will not be discussed further.
The adenine base of the first nucleotide of c-di-AMP is recognized specifically by LmPstA, through hydrogenbonds between its N1 and N6 atoms with the main chain amide and carbonyl of Gly47 (at the end of b3), respectively (Fig. 3B). One face of this adenine is packed against the Gly26-Ala27 amide bond (near the beginning of b2), while the other face is in contact with the side chains of Thr97 and Phe99 (in b4) from the neighboring monomer. The 2 0 OH of the ribose has hydrogen-bonding interactions with the side chains of Asn41 (b3 of neighboring monomer) and Thr28 (b2), with the latter also hydrogenbonded to N3 of adenine. The phosphate group is hydrogen-bonded to the side chain His108 (b5) and the main chain amide of Gly94 (B-loop of neighboring molecule).
The binding site for this nucleotide of c-di-AMP in LmPstA is equivalent to that for ATP in Mtb P II (Fig. 3C). In contrast the recognition of the second nucleotide of the c-di-AMP molecule is mediated by unique structural features in PstA, due primarily to the significant structural differences in the T-loop which forms a part of the intersubunit cleft where the ligand binds. The extended T-loop of Mtb P II would clash with the adenine of the second nucleotide of c-di-AMP (Fig. 3C). Instead, the T-loop in PstA forms a shorter, more structured loop with a conserved 34-GGFL-37 sequence in the turn. The main chain amides of Gly35, Phe36, and Leu37 have hydrogen-bonding interactions with the phosphate group of this nucleotide. The side chain of Phe36 is p-stacked against one face of the adenine, while its other face is exposed to the solvent.
In Mtb P II structure, the B-loop and C-loop are important for anchoring the triphosphate moiety of ATP (Fig. 3C). The B-loop has a highly conserved Walker Alike sequence motif (TGxxGDGKI) (Huergo et al. 2013) which interacts with the ATP triphosphate, and the Cloop also interacts with the ATP triphosphate through several arginine residues. In sharp contrast, PstA lacks the C-loop entirely, and the B-loop has a 30-residue flexible extension instead of the shorter, more structured B-loop found in P II proteins. In addition, the B-loop of PstA does not contain a Walker A-like motif. As a result, the band c-phosphates of ATP are unlikely to have strong interactions with PstA, which may explain why the protein does not bind this nucleotide.
To better characterize the c-di-AMP binding site in PstA, residues that were found to interact with c-di-AMP were substituted to alanine (Fig. 3D). In particular, we mutated several conserved residues on the T-loop including Phe36 and Asn41. The F36A mutant exhibited reduced binding to c-di-AMP with a K d of 7 lmol/L. The N41A mutant had even weaker binding to c-di-AMP with a K d of~140 lmol/L, though saturation was not achieved in this binding titration, precluding an accurate measure of the affinity. These results confirm that the c-di-AMP binding site identified by structural analysis is the true binding site in solution.

Conformations of the c-di-AMP molecules
The adenine base of the first nucleotide in all three c-di-AMP molecules is in the anti conformation (Fig. 3B), as is the case with ATP bound to P II . This is consistent with the several specific interactions the adenine has with PstA. In comparison, the adenine base for the second nucleotide in two of the c-di-AMPs assumes a syn conformation while the third one has an anti conformation ( Fig. 4A and B). In both conformations, the adenine base is pstacked against the Phe36 side chain. The ability of this adenine to adopt syn and anti conformations is consistent with the relative weak interactions it has with PstA. It is unclear if this conformational flexibility has any functional relevance.
In the structure of the ydaO riboswitch in complex with c-di-AMP, the c-di-AMP molecules also have one adenine in syn and one in anti (Gao and Serganov 2014; Jones and Ferre-D'Amare 2014; Ren and Patel 2014). All the adenines in the anti conformation have the C3 0 endo pucker for its associated ribose, while those in the syn conformation have a C2 0 endo ribose pucker (Fig. 4C).

Structural changes in PstA upon c-di-AMP binding
The overall structure of the c-di-AMP-bound form of the LmPstA trimer is almost identical to the apo structure with an r.m.s. distance of 0.6 A for their Ca atoms when they are overlaid (Fig. 5A). However, there are significant differences in the conformations of the T-loop and Bloop between the apo and c-di-AMP-bound structures. In the apo LmPstA structure, the T-loop from residues 33-39 are largely disordered, indicating that this loop is flexible in the absence of c-di-AMP binding. In addition, the B-loop kinks at a conserved 94-GGA-96 motif upon c-di-AMP binding significantly change the location and orientation of the loop (Fig. 5B, Video S1). Several interactions between LmPstA and c-di-AMP mediate the movement of the B-loop upon ligand binding. The main chain nitrogen of Gly94 interacts with the phosphate of the first nucleotide of c-di-AMP, and the main chain carbonyl of Gln92 interacts with the 2 0 OH of the second nucleotide.
The C-terminal His-tag of the apo LmPstA structure forms b-sheet interactions with the B-loop, possibly affecting its conformation. Thus, we also compared the cdi-AMP-bound LmPstA with the apo form of a PstA homolog from P. pentosaceus which was crystallized with an N-terminal His-tag (Fig. 5C). The same differences in the B-loop are also seen in this comparison, and because the B-loops are more ordered in the P. pentosaceus structure, the conformational changes that occur upon c-di-AMP binding are even more apparent. While our article was under preparation and review, the structures of PstA from S. aureus (SaPstA) (Campeotto et al. 2015;Muller et al. 2015) and a PstA homolog DarA from Bacillus subtilis (Gundlach et al. 2015) bound to c-di-AMP were reported. The detailed interactions with c-di-AMP for SaPstA and DarA are largely identical to what we observed for LmPstA. The main difference in the ligand binding mode between these structures and LmP-stA is that the second adenine of c-di-AMP bound to SaPstA and DarA is sandwiched between the side chains of Phe36 and Arg26, forming p-stacking and cation-p interactions. This Arg26 residue is present in most PstA proteins (Fig. 1A), but in LmPstA this residue is instead a glycine and the cation-p interaction is absent. This indicates that the arginine side chain is dispensable for cdi-AMP binding. In addition, the c-di-AMP molecules in the SaPstA and DarA structures all adopt an anti conformation, in contrast to the LmPstA structure. The presence of the Arg26 residue may promote the anti conformation of the adenine base.
The conformational change of the B-loop upon c-di-AMP binding that was observed for LmPstA is also present in DarA, but is absent in the Muller et al. (2015) SaPstA structure. The Campeotto et al. (2015) SaPstA structure shows one B-loop with a conformational change, while this change is absent in the other two B-loops. These differences in the B-loop conformations may be due to crystal-packing influence in the SaPstA structures, in which the N-terminal His-tag of an adjacent SaPstA trimer makes several interactions with the c-di-AMP molecule in place of the B-loop.

Functional studies of PstA
To interrogate the role of PstA-mediated c-di-AMP signaling in L. monocytogenes, we generated ΔpstA clean deletions in both WT and a conditional depletion strain of the sole diadenylate cyclase dacA (cΔdacA) (Witte et al. 2013) (Tables 3 and 4). Previous studies have shown that lowering the c-di-AMP level in L. monocytogenes causes a severe growth defect in nutritive media, defects in intracellular growth, and enhanced bacteriolysis in these environments. The absence of pstA in the cΔdacA strain resulted in a slightly improved growth rate, with the cΔdacA ΔpstA mutant exhibiting a 70-min doubling time, while cΔdacA exhibited an 80-min doubling time. The ΔpstA mutant had no effect in the WT parental background (Fig. 6A). Bacteriolysis measured as b-Gal release into the culture supernatant also revealed a partial rescue of the cell lysis defect in the cΔdacA strain, with the double cΔdacA-ΔpstA mutant exhibiting a twofold reduction relative to the parental strain alone (Fig. 6B). Introduction of pstA on a plasmid with a constitutively active promoter (P SPAC ) resulted in enhanced broth lysis beyond levels observed in the cΔdacA strain, likely as a result of over complementa-tion of PstA levels in this strain. In contrast to our expectation, there was no change in cefuroxime sensitivity for cΔdacA strain after deletion of pstA (Fig. 6C), suggesting that the stabilization of bacterial cell after pstA deletion in the cΔdacA strain may not be directly linked to a weakened cell wall.
We next assessed the effect of PstA on intracellular growth in macrophages and fibroblasts. In accordance with the broth growth, deletion of pstA had no effect on the growth of the WT strain but resulted in a small but reproducible effect on the growth defect exerted by loss of c-di-AMP production. Specifically, the double cΔdacA-ΔpstA mutant was observed to reach slightly higher levels of intracellular growth after 8 hours of BMM infection (Fig. 6D). To examine the effect of PstA over an extended infection period, fibroblast plaque assays were performed for 3 days. In accordance with macrophage growth, the plaques formed by the double cΔdacA-ΔpstA exhibited a twofold increase in area relative to the cΔdacA strain (Fig. 6E). We previously showed that the cΔdacA mutant undergoes enhanced lysis in macrophages (Witte et al. 2013;Sureka et al. 2014). Therefore, we tested the intracellular lysis of pstA mutants by a DNA delivery assay (Sauer et al. 2010). The results showed a twofold reduction in cell lysis in the cΔdacA-ΔpstA double mutant compared to cΔdacA, whereas the ΔpstA remained unaffected (Fig. 6F). Together these studies suggest that in the absence of c-di-AMP production, PstA has a negative impact on bacterial growth and survival both in and out of the host and that these are likely due to effects on bacterial cell stability.

Sequence conservation
Sequence conservation analysis of the PstA protein family shows that the residues identified in the structure as being important for c-di-AMP recognition are highly conserved (Fig. 7A). These include the 34-GGFL-37 residues on the T-loop which interact with the second nucleotide of c-di-AMP. In addition, the 94-GGA-96 sequence at the B-loop kink which allows for the loop to move toward c-di-AMP is also universally conserved. The surface of the rest of the PstA trimer is less conserved with the exception of several residues on the bottom of the trimer which are likely important for the structural integrity of the trimer ( Fig. 7B and C). Unlike P II proteins where the extended T-loop is highly conserved (Huergo et al. 2013), the extended B-loop in PstA has only weak sequence conservation ( Fig. 7B and C). Notably, Tyr51 on the T-loop of P II proteins is strictly conserved and can be modified by reversible uridylylation or adenylation in some bacterial species (Huergo et al. 2013). P II proteins interact with their protein targets primarily through T-loop-mediated interactions, and the uridyl-modification of Tyr51 has been shown to disrupt the interaction of the P II protein GlnK with the ammonium channel AmtB (Conroy et al. 2007). The two tyrosine residues on the B-loop of PstA have only~70% sequence conservation (Fig. 1A). Furthermore, L. monocytogenes does not contain a homolog of the E. coli uridylyl-transferase GlnD. However, we did note that within the shotgun sequencing that previously identified LmPstA as a c-di-AMP binding protein (Sureka et al. 2014), the peptides mapping to the B-loop of LmPstA were notably absent (Fig. 7D), consistent with a modification to this region of the protein that precluded its identification. Together these observations provide intriguing albeit speculative evidence of a modification to this site. However, given the absence of GlnD homologs in the Firmicutes (Huergo et al. 2013), it is unlikely that if present this modification is uridylylation or adenylation of the B-loop tyrosine residues common to GlnD-mediated P II modification and may be due to phosphorylation, as has been described in cyanobacteria (Forchhammer and Tandeau de Marsac 1995). The core structure of P II and PstA are remarkably similar (r.m.s. distance 0.9 A for the Ca atoms) despite onlỹ 15-20% sequence identity. In addition, the recognition of the equivalent adenine of c-di-AMP and ATP are almost identical in PstA and P II proteins. The structures differ in the loop regions which are less important for the overall structural integrity of the trimer but are critical for defining the ligand binding site. Thus, it is likely that PstA and P II arose from a common protein ancestor, and diverged through evolutionary changes in these loop regions. On the basis of these observations, we conclude that PstA is related to the larger P II protein superfamily, but has diverged through evolution to specifically recognize the bacterial second messenger c-di-AMP and thus is structurally and functionally distinct from the canonical P II proteins.

Possible functions of PstA
PstA currently has no known function. Based on its structural similarity to P II proteins, it is likely that PstA functions through a P II -like mechanism of binding and modifying the activities of its protein targets. There have been several reports of P II proteins bound to various macromolecules. The stoichiometry of these complexes are either one P II trimer bound to a trimeric protein in the case of AmtB and N-acetylglutamate kinase (Conroy et al. 2007;Gruswitz et al. 2007;Llacer et al. 2007;Mizuno et al. 2007), or to three separate monomers in the case of PipX and DraG (Llacer et al. 2010;Rajendran et al. 2011). In almost all of these complexes, the T-loop plays a critical role in complex formation, and is found in a different conformation in each complex. A notable exception is the P II -DraG complex where the T-loop is not involved in binding DraG, which instead binds to the lateral face of the P II trimer close to the ADP binding site. In the case of AmtB and N-acetylglutamate kinase, ATP binding to P II promotes complex formation through a mechanism involving changes in the T-loop conformation (Durand and Merrick 2006;Llacer et al. 2007). In Mtb P II , the T-loop makes several contacts with ATP, supporting such a mechanism (Shetty et al. 2010). Thus, ligand binding induces a particular shape and orientation of the T-loop in P II which plays a key role in complex formation with its protein target.
Although we provide indications that PstA contributes to bacterial cell stability in the absence of c-di-AMP production, the identity of protein regulatory targets, which parts of the PstA trimer are involved in target recognition, and the cellular consequences of target recognition remain to be discovered. The genetic context of c-di-AMP riboswitches and the posttranslational effects on protein targets support a general role for c-di-AMP in osmotic stress responses in Firmicutes, within which PstA homologs are conserved Nelson et al. 2013;Sureka et al. 2014). Consistent with these observations, our findings link PstA-mediated c-di-AMP signaling to bacteriolysis. If canonical P II protein regulatory mechanisms provide any insight into the biological role of PstA, it is feasible that PstA may regulate membrane transporter function analogous to GlnK regulation of the ammonium transporter. Such transporters may be linked to ion or osmolyte levels that help to stabilize the cell. Conversely, metabolic effects reminiscent of P II regulation of glutamine synthetase may impose changes in cell wall synthesizing capacity or osmolyte synthesis to affect cell stability. Based on the large movement of the B-loop upon c-di-AMP binding, which dramatically changes the overall shape of the PstA trimer, it is likely that c-di-AMP binding can modulate the binding of PstA to its target proteins. However, the B-loop is significantly longer and less conserved than the P II T-loop, so any protein-protein interactions mediated by the B-loop are likely to be different than what has been observed for P II proteins. Clearly, identification and characterization of the binding partners of PstA will reveal significant insight into the mechanism by which PstA-mediated c-di-AMP signaling affects the physiology and cellular stability within the Firmicutes.