Solution Structure of the Guanine Nucleotide-binding STAS Domain of SLC26-related SulP Protein Rv1739c from Mycobacterium tuberculosis*

The structure and intrinsic activities of conserved STAS domains of the ubiquitous SulP/SLC26 anion transporter superfamily have until recently remained unknown. Here we report the heteronuclear, multidimensional NMR spectroscopy solution structure of the STAS domain from the SulP/SLC26 putative anion transporter Rv1739c of Mycobacterium tuberculosis. The 0.87-Å root mean square deviation structure revealed a four-stranded β-sheet with five interspersed α-helices, resembling the anti-σ factor antagonist fold. Rv1739c STAS was shown to be a guanine nucleotide-binding protein, as revealed by nucleotide-dependent quench of intrinsic STAS fluorescence and photoaffinity labeling. NMR chemical shift perturbation analysis partnered with in silico docking calculations identified solvent-exposed STAS residues involved in nucleotide binding. Rv1739c STAS was not an in vitro substrate of mycobacterial kinases or anti-σ factors. These results demonstrate that Rv1739c STAS binds guanine nucleotides at physiological concentrations and undergoes a ligand-induced conformational change but, unlike anti-σ factor antagonists, may not mediate signals via phosphorylation.

The structure and intrinsic activities of conserved STAS domains of the ubiquitous SulP/SLC26 anion transporter superfamily have until recently remained unknown. Here we report the heteronuclear, multidimensional NMR spectroscopy solution structure of the STAS domain from the SulP/SLC26 putative anion transporter Rv1739c of Mycobacterium tuberculosis. The 0.87-Å root mean square deviation structure revealed a four-stranded ␤-sheet with five interspersed ␣-helices, resembling the anti-factor antagonist fold. Rv1739c STAS was shown to be a guanine nucleotide-binding protein, as revealed by nucleotide-dependent quench of intrinsic STAS fluorescence and photoaffinity labeling. NMR chemical shift perturbation analysis partnered with in silico docking calculations identified solvent-exposed STAS residues involved in nucleotide binding. Rv1739c STAS was not an in vitro substrate of mycobacterial kinases or anti-factors. These results demonstrate that Rv1739c STAS binds guanine nucleotides at physiological concentrations and undergoes a ligand-induced conformational change but, unlike anti-factor antagonists, may not mediate signals via phosphorylation.
The majority of bacterial SulP proteins and all known eukaryotic SulP/SLC26 proteins carry a C-terminal cytoplasmic "sulfate transporter antifactor antagonist" (STAS) 4 domain that shares distant sequence homology with bacterial antifactor antagonist proteins (19,20). Bacterial factors control specific transcriptional programs through activation of core RNA polymerase. All factors are negatively regulated by anti-factors. In Bacillus subtilis, F ( F ) initiates and controls the sporulation program. The antifactor for F is SpoIIAB, a Ser/Thr kinase under the inhibitory control of the anti-antagonist SpoIIAA. ATP-liganded SpoII AB binds and inhibits F . SpoIIAA activates F through sequestration of SpoIIAB (ATP), which displaces F from its complex with SpoIIAB. SpoIIAA phosphorylation (at Ser 58 ) by bound SpoIIAB leads to its dissociation from the ADP form of SpoIIAB. Phosphorylated SpoIIAA can then be dephosphorylated by phosphatase SpoIIE (summarized in Ref. 21). However, the GTP-binding and hydrolase activities of SpoIIAA (22) have unclear physiological functions, and no known relationship to either sporulation or the growth-promoting functions of the ribosome-associated GTPase, Obg (23).
The structure of SpoIIAA has been solved by heteronuclear NMR in solution (24), and x-ray crystallography (25) in phosphorylated and unphosphorylated forms, and in complex with SpoIIAB bound to either ADP or ATP (21). Published prelim-inary structures of additional STAS domain proteins include the NMR solution structure of Thermotoga maritima putative antiantagonist TM1442 in phosphorylated and unphosphorylated states (26), the NMR and crystal structures of T. maritima putative anti-antagonist TM1081 (27), and the crystal structure of the putative stressosome component RsbS from Moorella thermoacetica (28). The SpoIIAA structure has been used to model the eukaryotic SLC26 STAS domain, including congenital chloride diarrhea missense mutations in the human SLC26A3 domain that result in subtle chemical shift changes as detected by 1 H-15 N two-dimensional HSQC NMR (29). However, the amino acid assignments of this structure have not been reported and, until recently, no other structural information had been released for STAS domains of SulP/SLC26 anion transporters.
We have overexpressed and purified the 560-aa putative SO 4 2Ϫ transporter Rv1739c from M. tuberculosis, comprising an N-terminal trans-membrane domain (aa 1-414) and a Cterminal cytoplasmic STAS domain (aa 442-560) (15). Previously, we reported the secondary structural features present in the C-terminal STAS domain using heteronuclear NMR spectroscopy (30). Here we report the three-dimensional NMR structure of the Rv1739c STAS domain (aa 437-560). Using intrinsic fluorescence quench data and 1 H-15 N chemical shift perturbation (CSP) experiments, we demonstrate that the monomeric STAS domain binds guanine nucleotides at physiological concentrations. The purified STAS domain is also associated with detectable GTPase activity. CSP analysis reveals that STAS binding of GDP elicits a more extensive conformational perturbation than does GTP. In silico docking calculations support these CSP data, suggesting that GDP binds preferentially, whereas GTP binding involves a subset of the GDP-interacting residues. The overexpressed Rv1739c STAS domain is not a phosphoprotein as isolated from E. coli or Mycobacterium smegmatis, and is not phosphorylated in vitro by recombinant Ser-Thr kinases or antifactors of M. tuberculosis. While this work was in editorial review, reports appeared presenting crystal and solution structures of a centrally deleted core portion of the STAS domain from rat Slc26a5/prestin (31) and a crystal structure of the full-length STAS domain from E. coli SulP protein YchM (18).

EXPERIMENTAL PROCEDURES
Purification of Rv1739c STAS Domain-A polypeptide encompassing aa 437-560 of Rv1739c was overexpressed for study of the STAS domain, based on alignment with STAS domains from E. coli ychM and Sultr1.2, and with structurally characterized STAS domains from antifactor antagonists of B. subtilis, Bacillus sphericus, and T. maritima 1442. For clarity, Rv1739c 437-560 are referred to as Rv1739c STAS aa 1-124 in this study. The Rv1739c STAS domain containing a C-terminal His 6 tag was overexpressed in the E. coli strain Tuner DE4pLacI (Novagen) as previously described (30). The protein purified from the nickel-nitrilotriacetic acid column was Ͼ96% pure as detected by Coomassie Blue staining of SDS-PAGE or PFO-PAGE gels. Subsequent gel filtration (Superdex 75 FPLC) yielded a homogeneous single peak eluting at a retention time that corresponds to monomeric STAS (ϳ15 kDa), and is observed as a single band on overloaded SDS-PAGE (supplemental Fig. S1). Protein identity was confirmed by mass spectrometry. Unlabeled and uniformly 13 C/ 15 N-or 15 N-labeled STAS-His 6 NMR samples of 0.7-1.0 mM concentration in 50 mM sodium phosphate, 275 mM NaCl, pH 7.2, were prepared and subjected to NMR spectroscopy (supplemental "Methods" and Ref. 30). Rv1739c STAS expression in M. smegmatis is described under supplemental "Methods".
NMR Spectroscopy-NMR experiments were performed at 298 K on a Bruker Avance 600 MHz spectrometer equipped with a 5-mm triple resonance PFG (z axis) probe. All NMR data were acquired in gradient-selected, sensitivity-enhanced mode. NMR data were processed using NMRPipe/NMRDraw processing software (32), and analyzed using ANSIG (33). Protein backbone and side chain resonances of 1 H, 13 C, and 15 N were assigned using standard two-dimensional, threedimensional, and triple resonance experiments (34 -36) as detailed under supplemental "Methods".
Structure quality was assessed with PROCHECK-NMR (41). The 25 structures of the highest refinement represented the solution structure of Rv1739c STAS. Figures were generated with MOLMOL (42) and VMD (43), and sequences were aligned with ClustalW and LALIGN.
In Silico Docking of STAS-Nucleotide Complexes-In silico docking calculations were carried out using AutoDock Vina (44). An exhaustive search was carried out to find and cluster the best docked poses of the GDP and GTP complexes with Rv1739c STAS. The average structure of the STAS domain was used as a starting target molecule for these docking studies (see supplemental "Methods").
Photoaffinity Labeling of Rv1739c STAS Domain-Rv1739c STAS was incubated with biotinylated azido-GTP in the presence or absence of 2 mM GTP, and briefly irradiated. An aliquot of the terminated reaction was electrophoresed, blotted, developed, and visualized as detailed (supplemental "Methods").
Intrinsic Fluorescence Quench of Rv1739C STAS-Gel filtration-purified Rv1739c STAS in buffer A (25 mM Tris, 100 mM NaCl, pH 7.2) was centrifuged before use. Free acid forms of ATP, GTP, and GDP (Sigma), GTP␥S, and GDP␤S (Jena Biosciences) of the highest purity available were prepared as stock solutions in buffer A. Rv1739c STAS (12-20 M; initial volume 200 l) was preincubated in 48-well plates at 24°C with nucleotides in the absence or presence of 1 mM added MgCl 2 (final volume: 217 l). Steady-state intrinsic fluorescence intensity of RV1739c STAS was then recorded at em 290 -400 nm at 2-nm intervals with fixed ex ϭ 280 nm (SpectraMax M5, Molecular Devices, Sunnyvale, CA) in the presence of sequentially increasing nucleotide concentrations or titrated with buffer A alone. Rv1739c STAS fluorescence at each nucleotide concentration was corrected for dilution and for inner filter effect contributions of added nucleotide, as described (45). The effects of identical sequential increases in added nucleotide were assessed on the intrinsic fluorescence of a mixture of free tryptophan plus 3 molar eq of free tyrosine (the 1:3 Trp:Tyr molar ratio reflected in the native Rv1739c STAS amino acid sequence), with identical data acquisition parameters and instrument settings. The inner filter effect correction factor was determined as the ratio of fluorescence intensity of the Trp:Tyr solution in the absence of nucleotide to that in the presence of nucleotide at each emission wavelength. The inner filter effect correction factor was Յ2 at all nucleotide concentrations.
Inner filter effect-corrected, normalized fluorescence intensity data were plotted as a function of nucleotide concentration and fit (SigmaPlot, Systat) to a single site ligand-binding model, where F 0 Ϫ F is the change in fluorescence intensity at 316 nm ( max ) following the addition of nucleotides at concentration [S], ⌬F max is the maximum change in fluorescence intensity, and K d is the STAS-nucleotide dissociation constant. Assay of Nucleotidase Activity-Rv1739c STAS or human RhoA-GST (as a positive control) were mixed with [␥-32 P]GTP or [␥-32 P]ATP. Reactions were stopped with SDS, and fractionated by thin layer chromatography (TLC). Excised spots containing inorganic [ 32 P]phosphate were counted (Packard Tri-Carb 2200CA) in Ecoscint A (National Diagnostics, Atlanta, GA) with 97% efficiency (supplemental "Methods").

Distance and Dihedral Restraints of Rv1739c STAS Domain
NMR-derived experimental restraints used in determining the three-dimensional structure of Rv1739c STAS consisted of 1692 unique NOEs, 88 hydrogen bond restraints, and 163 (, )-backbone angle restraints (Table 1). Sequential and medium range NOEs are summarized in Fig. 1. Of 311 long range NOEs assigned, 68 are from 1 H ␣ and 1 H ⌵ backbone-side chain interactions, 200 are from side chain-side chain interactions, and 43 represent backbone-backbone interactions. Hydrophobic and aromatic residues contribute 62 and 34% of the long range NOEs identified in Rv1739c STAS.  Fig. 2A illustrates an ensemble of the 25 best NMR conformers of the Rv1739c STAS domain superimposed on their backbone N, C ␣ , and CЈ atoms. These structures were selected using standard structural criteria including: low target function, minimal upper distance NOE violations Ͼ0.5 Å, and absence of dihedral angle violations Ͼ5°. The low root mean square deviation values represented by the superimposed ensemble and the identification in allowed regions of the Ramachandran map of 98.1% of amino acid residues support the precision and stereochemical quality of the calculated structures (Table 1).     NMR structure of T. maritima TM1442, and the crystal structure of Rattus norvegicus Slc26a5/prestin. The structures are quite similar despite divergent primary sequence identities of Ͻ18%.

Structural Comparison of Rv1739c STAS with Other STAS Domains
The overall tertiary fold of Rv1739c STAS resembles that of SpoIIAA from B. sphaericus and unphosphorylated SpoIIAA from B. subtilis, with one face of the central ␤-sheet juxtaposed to helices ␣1 and ␣2, and the other face exposed to the aqueous environment. The flexible structure of the N-terminal 14 aa residues of Rv1739c STAS contrasts with the structured N-terminal regions of SpoIIAA from B. sphaericus and B. subtilis. These include single ␤ strands of 8 or 3 residues, respectively, packed in antiparallel orientation to the subsequent ␤ strands arrayed in parallel, which together form a central ␤ sheet. Helices ␣1 and ␣2 of Rv1739c STAS and B. subtilis SpoIIAA are similar in length, but shorter than the corresponding helices in B. sphaericus SpoIIAA. Helix ␣4 of Rv1739c STAS is oriented away from the protein core as in B. sphaericus, but closely packed to the main fold in B. subtilis and TM1442. Rv1739c STAS helix ␣3 extends parallel to the central ␤-sheet, in contrast to a perpendicular orientation of ␣3 in B. subtilis SpoIIAA. ␤1 and ␤2 strands of Rv1739c STAS and TM1442 are of similar length, but shorter than those of SpoIIAA. The ␤3-strands in Rv1739c (aa 80 -86) and B. sphaericus SpoIIAA are similar in length. However, this region is shorter in length in TM1442 and unstructured in B. subtilis SpoIIAA.
The topological fold of Rv1739c STAS closely resembles that of the rat prestin construct truncated at its C terminus and de-void of its "intervening sequence" (IVS) region aa 564 -636. A notable exception to this similarity is the N-terminal 15-aa extension present in prestin STAS, which includes a "␤1 strand" at its extreme N terminus and a "␤0 strand" in the immediately juxtaposed loop (31). Structural comparison among SulP/SLC26 polypeptides reveals that the presence of a common C-terminal ␤4 strand (aa 110 -111 of Rv1739c STAS) is often absent from antifactor antagonist structures. In addition, whereas helices ␣1 and ␣2 of anti-factor antagonists are arrayed in parallel orientation, the corresponding ␣1 and ␣2 helices of SulP/SLC26 STAS domains are oriented along divergent axes.
The conserved 56 DSSG motif of SpoIIAA constitutes a canonical phosphorylation site at Ser 58 (Fig. 3E, in red). Sequence alignment suggests that Thr 64 of Rv1739c, 62 DLTA (also in red), provides a homologous phosphorylation site of Rv1739c. Superimposed tertiary structures further support the alignment of Ser 58 of SPOIIAA (PDB code 1AUZ) with the partially surface exposed Thr 64 of Rv1739c STAS (PDB code 2KLN). These phosphorylation motifs in both SpoIIAA and Rv1739c are bracketed by hydrophobic residues. In addition to Thr 64 , many conserved residues of Rv1739c STAS reside within a localized patch comprising the ␤2-␣2 region (Fig. 3E).

Guanine Nucleotide Binding by Rv1739c STAS
The reported nucleotide binding properties of the antifactor antagonist SpoIIAA (22) and the STAS domain of the blue light receptor YtvA from B. subtilis (47) prompted us to evaluate the nucleotide binding potential of Rv1739c STAS.
As illustrated in Fig. 4A, Rv1739c STAS was photolabeled by a ␥-biotinylated derivative of 8-N 3 -GTP, and this photolabeling was prevented in the presence of excess unlabeled GTP. The presence of a unique Trp residue, Trp 50 (aa 486 of Rv1739c holoprotein), in Rv1739c STAS and Tyr residues at STAS positions 7, 19, and 21 ( Fig. 3E and supplemental Fig. S4A) allowed assessment of nucleotide binding using intrinsic fluorescence quench methods.
Rv1739c STAS fluorescence emission max was observed at 316 nm. Fig. 4B illustrates the gradual quench of intrinsic STAS fluorescence with progressively increasing concentrations of GDP. Supplemental Fig. S4 shows comparable quench by increasing concentrations of GTP, of non-hydrolyzable guanine nucleotides, and ATP. The observed quench isotherms are compatible with a single nucleotide binding site with dissociation constants (K1 ⁄ 2 , in M) of 146 Ϯ 38 for GDP, 347 Ϯ 227 for GTP (Fig. 4, C and D), 152 Ϯ 53 for GDP␤S (supplemental Fig. S4, E and F), and 326 Ϯ 131 for GTP␥S (supplemental Fig. S4, C and D and Table 2). K1 ⁄ 2 values (in M) were higher in the presence of 1 mM MgCl 2 , 587 Ϯ 337 for GDP and 727 Ϯ 251 for GTP (Table 2). K1 ⁄ 2 for ATP in the absence of MgCl 2 was 756 Ϯ 232 M (supplemental Fig. S4, G and H). Thus, the nucleotide affinity rank-order profile for Rv1739c STAS appeared to be GDP Ͼ GTP Ͼ ATP. Maximum STAS fluorescence quench was observed at Յ400 M nucleotide except in the case of ATP, which required 2 mM nucleotide. The percent quench varied from 19 (for GTP) and 49% for GDP (Table 2). These data suggest a nucleotide interaction surface in Rv1739C STAS that directly or indirectly perturbs Trp 50 and/or the three Tyr residues through conformational effects on adjacent or interposed residues.

GTPase Activity Associated with Rv1739c STAS
Intrinsic signaling or enzymatic activities of bacterial SulPassociated STAS domains remain undefined. B. subtilis SpoIIAA exhibits GTPase activity (22), but the GTP-binding protein YtvA lacks detectable GTPase activity (48). We therefore evaluated Rv1739c for the presence of intrinsic nucleotidase activity. As shown in supplemental Fig. S5A, the FPLCpurified Rv1739c STAS preparation exhibited modest GTPase activity that was STAS-dependent, temperature-sensitive, inactivated by SDS (supplemental Fig. S5B), and time-dependent. STAS-associated ATPase activity was not detected (not shown).

Fig. 5 shows two-dimensional 1 H-15 N HSQC overlays of
Rv1739c STAS domain resonances in the absence (blue contours) and presence (red contours) of saturating concentrations of GTP (A) or GDP (B). Perturbations with weighted CSPs (⌬␦ weighted ) Ն 0.05 were considered significant (supplemental Fig. S6). CSP analysis revealed that the GDP-induced conformational change in Rv1739c STAS (Fig. 5B) affected a larger number of residues with greater extent of perturbation than those altered by the addition of GTP (Fig. 5A). GDP significantly perturbed 16 residues (red in Fig. 6, C and D), including Ala 10 , Arg 12 , Val 13 , Gly 15 , Val 41 , Asp 44 , Gln 47 , Val 48 , Arg 79 , Gly 106 , Glu 107 , Asp 108 , His 109 , Ile 110 , Arg 122 , and Arg 124 . GTP perturbed a subset of these residues (red in Fig.  6, A and B), including Ala 10 , Arg 12 , Val 13 , Gly 15 , Val 48 , Val 61 , Arg 79 , His 109 , and Arg 124 . The results are consistent with the lower K1 ⁄ 2 value for GDP than for GTP ( Fig. 4 and Table 2). The amino acids exhibiting the largest GDP-induced CSPs are illustrated at higher resolution in supplemental Fig. S6, A and  B.
The observed nucleotide-induced conformational perturbations were predominantly localized within the solvent-exposed loops and immediately adjacent residues, as evidenced by the conformational flux of Val 41 at the end of helix ␣ 1 , of Ile 110 localized at the end of strand ␤ 4 , and Arg 122 near the end of the C-terminal helix ␣ 5 . Many nucleotide-perturbed residues localize to a single face of Rv1739c STAS. Val 61 , located near the semi-conserved, potential hydrolase nucleophile Thr 64 , is the only residue perturbed Ͼ0.05 ppm by GTP alone (Fig. 6A and supplemental Fig. S6C).
Addition of GTP or GDP to Rv1739c resulted in emergence in the 1 H-15 N two-dimensional HSQC spectra of additional resonances (circled in Fig. 5, A and B) of near-identical ( 1 H, 15 N) coordinates. Line widths were broadened to a greater degree by GDP than by GTP. These ligand-induced resonances may belong to previously unassigned loop residues and/or to Ser 95 in helix ␣3 (30). Alternatively, these resonances may reflect slow conformational exchange of select residues.

In Silico Docking of the STAS-Nucleotide Complex
Docking of GDP and GTP into the three-dimensional structure of Rv1739c STAS was performed using AutoDock Vina (44). These in silico experiments identified two potential nucleotide binding sites in Rv1739c STAS (supplemental Fig.  S7). One site involves residues localized primarily near the STAS N terminus, whereas the other includes predominantly C-terminal residues.
STAS-GDP Docking-A flexible grid-based GDP-STAS docking protocol defined 9 clustered poses for the STAS domain bound to GDP, as illustrated under supplemental Fig.   S7. In the lowest energy pose, the GDP-interacting amino acid residues include Gln 9 , Gly 15 , Tyr 19 , Leu 77 , Arg 78 , Arg 79 , Gly 80 , Lys 104 , Ile 105 , Glu 107 , Asp 108 , Pro 115 , Val 118 , Gln 119 , and Arg 122 . Many of these residues are, or are adjacent to, those exhibiting above-threshold CSP in NMR titration studies. GDP binding is preferentially clustered to C-terminal STAS residues, with 7/9 poses supporting this preferred docked complex.

Rv1739c STAS Appears Not to Be a Phosphoprotein
Mass spectrometric analysis of the Rv1739c STAS domain isolated from E. coli revealed no phosphate. Because the Rv1739c STAS polypeptide may have undergone dephosphorylation during purification, we evaluated its ability to serve as an in vitro phosphorylation substrate. Supplemental Fig. S8 shows, however, that none of the 11 recombinant, purified M. tuberculosis Ser/Thr kinases stably phosphorylated the Rv1739c STAS domain in vitro. Moreover, Rv1739c STAS did not serve as the phosphorylation substrate for any of the four purified, recombinant M. tuberculosis antifactors.
Given that the isolated Rv1739c STAS may be a poor phosphorylation substrate in vitro, we assessed the phosphorylation status of FLAG-tagged Rv1739c STAS overexpressed in M. smegmatis. Importantly the Rv1739c overexpressed in M. smegmatis lacked an immunodetectable phosphothreonine (not shown), further supporting the conclusion that Rv1739c is not a substrate for mycobacterial Ser/Thr or antifactor kinases.

DISCUSSION
We have reported the three-dimensional NMR structure of the STAS domain from the SulP/SLC26 polypeptide Rv1739c of M. tuberculosis. We have compared this structure with previously solved structures of anti-factor antagonists, as well as with structures of STAS domain structures from putative anion transporters reported during review of this manuscript. Rv1739c STAS exhibited binding of nucleotides with affinities at the upper end of the physiological range, and with rank order of GDP Ͼ GTP Ͼ Ͼ ATP. GDP and GTP elicited distinct conformational changes in the Rv1739c STAS structure. Low rates of GTP hydrolysis were associated with nominally pure Rv1739c STAS. Rv1739c STAS underwent no detectable phosphorylation in vitro, or when overexpressed in E. coli or M. smegmatis.
Solution Structure of Rv1739c STAS-Despite its very limited amino acid sequence similarity to other STAS domains, the Rv1739c STAS solution structure superimposes closely with previously reported NMR and x-ray crystallography structures of antifactor antagonists. However, in its absence of a ␤ strand within the most N-terminal residues 1-14, Rv1739c STAS differs from the structured N-terminal regions observed in the other anti-factor antagonist structures (see above). The lack of N-terminal structure may be intrinsic to Rv1739c STAS, perhaps due in part to the presence of two proline residues: Pro 8 and Pro 14 , which may contribute to the conformational heterogeneity and general disruption of short ␤ strand formation in this region. Alternatively, the lack of N-terminal structure may reflect the absence of particular additional residues in this construct (i.e. ϳ420 -426) that link the reported STAS domain structure to the C-terminal end of the transmembrane domain, or the absence of a physiological interacting protein or ligand. With the exception of this unstructured N-terminal region, the Rv1739c backbone is a well defined, rigid/ordered structure, as shown by heteronuclear NOE data (supplemental Fig. S3B). The short intervening loop connecting regions of the secondary structure in Rv1739c resemble those of other antifactor antagonists and E. coli SulP ychM (18), but contrast with the longer intervening loops in STAS domains of mammalian SLC26 proteins (Fig. 3, D and E).
The Rv1739c STAS domain construct studied here lacks the 15-aa N-terminal extension present in the crystal structure of the C-terminal truncated, IVS-deleted rat prestin STAS structure (Fig. 3E). This 15-aa extension includes the first ␤-strand ("␤1") observed in rat prestin STAS (31). Deletion from the prestin STAS domain of its 73-aa IVS, of yet unknown function, was crucial for its crystallization, as was truncation from the rat prestin construct of its C-terminal 9 aa (31). Indeed, the secondary structure prediction program Psipred also predicts the presence of this ␤1 strand in rat prestin STAS containing the IVS, but considerably shortened compared with ␤1 in the crystal structure. Moreover, Psipred predicts the presence of the ␤1 strand in all human SLC26 STAS domains modified to include comparable IVS deletions. However, without computational deletion of this IVS, only SLC26A1, SLC26A2, and SLC26A7 are predicted to retain the ␤1 strand, whereas SLC26A3, SLC26A4, SLC26A6, SLC26A8, SLC26A9, and SLC26A11 are predicted to lack the ␤1 strand.
In contrast, the secondary structure prediction program Yaspin suggests that the presence of this STAS ␤1 strand of prestin is IVS-independent in rat prestin and in some other human SLC26 polypeptide paralogs. However, Yaspin also predicts the absence of the ␤1 strand structure from the STAS domains of human SLC26A1, SLC26A3, SLC26A4, and SLC26A11, as well as from the STAS domains of Arabidopsis thaliana sulfate transporters Sultr2.1 and Sultr2.2. Thus, the ␤1 strand may not be an essential part of all STAS domain structures, and its presence in some SLC26 gene products may be influenced by the presence or absence of the IVS sequence.
Residues Pro 543 , Tyr 545 , Tyr 546 , and Val 655 of the rat prestin STAS domain have been proposed to constitute a juxtamembrane binding surface. The corresponding Rv1739c STAS residues Pro 24 , Cys 26 , Phe 27 , and Thr 64 are similarly oriented. Such an interaction could be with the lipid bilayer or, as proposed for Synechococcus BicA, with a cytoplasmic loop of the SulP transmembrane domain.
Rv1739c STAS Is a Nucleotide-binding Protein-Photoactivated 8-N 3 -biotinyl-GTP labeled Rv1739c STAS, and labeling was blocked in the presence of excess GTP (Fig. 4A). We monitored the intrinsic fluorescence of Rv1739c STAS as a function of added nucleotide rather than directly monitoring fluorophor-labeled nucleotides, because these latter experiments often exhibit unique, fluorophor-related changes in ligand binding affinity and/or hydrolytic activity that are specific to individual GTP-binding proteins (49). Both GTP and GDP quenched the intrinsic fluorescence of Rv1739c STAS (Fig.  4), consistent with nucleotide binding directly or indirectly altering the electrostatic environments of residues Trp 50 and/or Tyr 7 , Tyr 19 , or Tyr 21 . The K1 ⁄ 2 for GDP binding was 146 M, about half the value for GTP. K1 ⁄ 2 values for both GDP and GTP were indistinguishable from those of their respective thioester analogs. The K1 ⁄ 2 for binding of ATP was ϳ2-fold higher than for GTP (supplemental Fig. S4F), a property shared with SpoIIAA (22). The presence of 1 mM Mg 2ϩ reduced binding affinity 2-fold for GTP and ϳ4-fold for GDP (Table 2), as noted for the STAS domain of B. subtilis photosensor YtvA (48). These data are the first to demonstrate guanine nucleotide binding to a SulP transporter STAS domain. The low guanine nucleotide binding affinities of Rv1739c STAS are within the range of previously reported GTP binding constants as high as 1.1 mM (supplemental Table S1), but contrast with the GTP K1 ⁄ 2 of 0.25 M reported for B. subtilis SpoIIAA (22).
Rv1739c STAS Nucleotide Binding Site(s)-NMR CSP studies ( Fig. 5 and supplemental Fig. S6) suggest a semi-contiguous surface in Rv1739c STAS involved in binding GDP and/or GTP (Fig. 6). Among the Rv1739c STAS residues perturbed by Ͼ0.05 ppm in the presence of nucleotides, Val 41 and Val 61 (and Val 48 to a lesser degree) are hydrophobic residues conserved in other antifactor antagonists and in STAS domains (Fig. 3D). In silico docking of STAS with GDP and GTP (supplemental Fig. S7) supported involvement of residues per-turbed in the CSP studies (Figs. 5 and 6, and supplemental Fig. S6). Similar to the NMR CSP data, 7 of the 9 calculated poses in the presence of GDP support binding to an interface that involves residues localized within the C-terminal region of the STAS domain, including aa Gly 106 -Ile 110 (Fig. 6C and supplemental S6D). The CSP data suggest (with support from the docked poses) that the C-terminal binding site is perturbed to a greater degree by GDP than by GTP (Figs. 5 and 6, and supplemental Fig. S6, C and D). GTP docking involved only a subset of residues at this interface, as well as additional residues of the STAS N-terminal region (supplemental Fig.  S7A). Some of the nucleotide-perturbed residues, including those of the N-terminal region, Val 48 and Val 61 , correspond to regions in anti-factor antagonist TM1081, which exhibit millisecond scale conformational changes in solution but not in the crystal, suggesting their involvement in binding or catalytic activity (27).
The CSP data and docking calculations suggest two possible nucleotide binding sites in the presence of 20 mM nucleotide, whereas intrinsic fluorescence quench data at 30 -100-fold lower nucleotide concentrations are compatible with a single binding site ( Table 2). The single binding site predicted from fluorescence quench measurements may be one of the two sites suggested by CSP and docking calculations, with recruitment of the second site only at much higher ligand concentrations. Alternatively, both sites may bind nucleotide in the physiological concentration range, with only one of the sites influencing intrinsic fluorescence. The positioning of the Tyr and Trp residues in Rv1739c STAS suggests the N-terminal site could be that site.
GTPase Activity Associated with Rv1739c STAS-Antifactor antagonists differ in their nucleotide hydrolase activities. Whereas SpoIIAA hydrolyzed GTP (22), YtvA had undetectable GTP hydrolase activity (47,50). Rv1739c STAS-associated GTPase activity was detectable as slow inorganic [ 32 P]phosphate release from [␥-32 P]GTP (supplemental Fig.  S5) in the nominal absence of Mg 2ϩ . The hydrolysis rate was not increased upon addition of 1 mM Mg 2ϩ . Rv1739c STASassociated GTPase activity was only ϳ8-fold lower than that of RhoA-GST in the same conditions. The low rates thus likely reflect low input substrate concentrations, because measured Rv1739c STAS GTPase activity was orders of magnitude slower than that reported for SpoIIAA in the presence of the ratio [GTP]/K1 ⁄ 2(GTP) ϭ 30 (22). However, Rv1739c STAS-associated subsaturated GTPase activity was only ϳ20 -50-fold lower than the GTP-saturated GTPase rates of Saccharomyces cerevisiae Ypt6 (51, 52) and human Rheb (49). Thus, Rv1739c STAS-associated GTPase activity was near the low-end of the reported GTPase rates. Nonetheless, the presence of copurifying contaminant GTPase(s) in our Rv1739c STAS preparation remains a potential explanation.
Possible Physiological Roles of GTP Binding by Rv1739c STAS-The GTP K1 ⁄ 2 value within the intracellular range of [GTP] suggests a physiological role for nucleotide binding by Rv1739c STAS. Nucleotide binding might regulate Rv1739c localization or anion transport that is mediated by or associated with Rv1739c function, directly or through alteration of protein-protein interaction.
Rv1739c STAS, unlike SpoIIAA, was not detectably phosphorylated (supplemental Fig. S8). The mechanistic role of SpoIIAA phosphorylation remains unclear, because substitution of this serine residue by acidic residues did not phenocopy the activity of serine phosphorylation (24,25,27). The STAS domains of human SLC26A3 (29) and human SLC26A9 (53) interact in vitro with the phosphorylated R domain of CFTR, but the reported regulatory interactions of SLC26 polypeptides and CFTR in vitro and in vivo have varied. PKC phosphorylation of the human SLC26A6 STAS domain has been suggested to regulate Cl Ϫ /HCO 3 Ϫ exchange activity through CAII binding (54), but this phosphorylation did not regulate oxalate transport (55), and its relationship to CFTR is unknown. Although the R domain of CFTR is not shared among other ABC proteins, Rv1739c might function as a regulator of the M. tuberculosis ABC family sulfate permease CysTWA (15). E. coli ychM STAS domain has been shown to interact with subunits of other ABC family permeases (18).
M. bovis cysA mutants lack sulfate uptake in resting cultures (56), yet cysA and subI sulfate uptake mutants of Mycobacterium bovis survive in mice as well as wild type M. bovis (57). Thus, under the stressful conditions inside acidifying compartments of cells, mycobacteria might rely on stressregulated expression of SulP Rv1739c. Indeed, Rv1739c is among the sulfate transport and metabolism genes up-regulated in the dormant phase of infection in cultured macrophages and in granulomas of intact mice (58). Similarly, the STAS domain-containing SulP protein ychM of E. coli is also up-regulated in stress conditions of RpoN (54) deletion (59) and nitrofurantoin treatment (60), and severely down-regulated in a virulence-impaired E. coli O157:H7 strain with deletion of the conserved, alternative factor and stress regulator RpoS (61). The recent discoveries that E. coli ychM binds to acyl carrier protein, an abundant, small polypeptide with many additional interaction partners (62), and that its deletion influences multiple gene products that regulate fatty acid biosynthesis (18), add new complexity to these questions. Furthermore, stress regulation (as nutrient supply and redox potential) has been proposed to regulate, on a cyclical basis, the direct interaction of A. thaliana SulP sulfate transporter Sultr1;2 with the cysteine synthase, O-acetylserine (thiol) lyase (63). A STAS-related or STASmediated guanine nucleotide cycle could potentially regulate the proposed cycle of binding and unbinding. Thus, the STAS domain of SulP anion transporters may sense environmental or metabolic stress through nucleotide binding, exchange, and/or hydrolysis to modulate anion transport and/or scaffolding of signaling proteins or enzymes as part of the regulation of cell metabolism and other functions.