The domain architecture of PtkA, the first tyrosine kinase from Mycobacterium tuberculosis, differs from the conventional kinase architecture

The discovery that MptpA (low-molecular-weight protein tyrosine phosphatase A) from Mycobacterium tuberculosis (Mtb) has an essential role for Mtb virulence has motivated research of tyrosine-specific phosphorylation in Mtb and other pathogenic bacteria. The phosphatase activity of MptpA is regulated via phosphorylation on Tyr128 and Tyr129. Thus far, only a single tyrosine-specific kinase, protein-tyrosine kinase A (PtkA), encoded by the Rv2232 gene has been identified within the Mtb genome. MptpA undergoes phosphorylation by PtkA. PtkA is an atypical bacterial tyrosine kinase, as its sequence differs from the sequence consensus within this family. The lack of structural information on PtkA hampers the detailed characterization of the MptpA–PtkA interaction. Here, using NMR spectroscopy, we provide a detailed structural characterization of the PtkA architecture and describe its intra- and intermolecular interactions with MptpA. We found that PtkA's domain architecture differs from the conventional kinase architecture and is composed of two domains, the N-terminal highly flexible intrinsically disordered domain (IDDPtkA) and the C-terminal rigid kinase core domain (KCDPtkA). The interaction between the two domains, together with the structural model of the complex proposed in this study, reveal that the IDDPtkA is unstructured and highly dynamic, allowing for a “fly-casting-like” mechanism of transient interactions with the rigid KCDPtkA. This interaction modulates the accessibility of the KCDPtkA active site. In general, the structural and functional knowledge of PtkA gained in this study is crucial for understanding the MptpA–PtkA interactions, the catalytic mechanism, and the role of the kinase–phosphatase regulatory system in Mtb virulence.

Mycobacterium tuberculosis (Mtb) 2 (1) increasingly evolves multidrug-resistant (2) and extensive drug resistant (3) strains, causing epidemic problems on a global scale (4). New therapeutics against tuberculosis (TB) are thus urgently needed. The development of new inhibitors is challenging, especially if one considers that the requirements of anti-TB drugs to exhibit bactericidal and sterilizing activity are complex and that the molecular mechanisms responsible for the Mtb pathogenic regulation are not yet fully understood (5). Cellular pathways that trigger virulence (such as enzymes that support the mycobacterial dormancy, persistency, and drug resistance) are potential targets for the implementation of new anti-tubercular agents. In particular, proteins involved in the Mtb signaling network appear interesting.
Reversible phosphorylation is co-regulated by kinases and phosphatases and provides a basis for the signal transmission in multiple organisms (6). The genome of Mtb encodes 27 enzymes involved in the cellular cross-talk; 12 two-component systems (7), 11 eukaryotic-like serine-threonine protein kinases (STPKs, PknA-PknL), one serine-threonine phosphatase (8), and two protein-tyrosine phosphatases (MptpA and MptpB) (9) were identified that adjust the phosphorylation status of specific amino acids (His/Asp and Ser/Thr/Tyr) during diverse cellular processes. Despite the well-established significance of the protein phosphomodification in mycobacteria, its functional role in the pathogenicity has remained poorly understood (10,11). Several STPKs for post-translational modifications of other endogenous kinases have been identified in Mtb.
In Mtb, phosphorylation of threonine residues is highly abundant compared with serine phosphorylation (12). The first strong indication regarding the importance of pTyr modifications for Mtb pathogenicity followed from the discovery of the low-molecular weight MptpA (17.5 kDa) and its critical role for Mtb virulence (13). During infection, MptpA is secreted from the pathogenic mycobacteria into the cytosol of host macrophages and interferes with the host endogenous signaling pathway. MptpA catalyzes dephosphorylation of the human vacuolar protein sorting 33B (VPS33B) complex, which leads to suppression of the phagosome-lysosome fusion (14,15). The virulence factor MptpA escapes the complex structure of the Mtb cell wall, which is otherwise impermeable to many drugs, thus making it an attractive target for inhibitor development. However, the exact mechanism that modulates the activity of MptpA and its translocation from the bacterial cytosol into the cytosol of the infected macrophages is still unknown.
On the molecular level, our detailed structural characterization of MptpA indicated a conformational rearrangement from the open to a closed state, induced by ligand binding (16). One oftheproposedmechanismsofMptpAregulationinvolvesphosphorylation of the vicinally located tyrosine residues (Tyr 128 / Tyr 129 ) (17). To the best of our knowledge, there is no information available about the phosphorylation status of each tyrosine or on the regulation of the tyrosine phosphorylation. Until now, only one protein-tyrosine kinase (PtkA, 30.6 kDa) has been found in Mtb (18). PtkA was classified as an atypical bacterial kinase, or so-called "odd" tyrosine kinase (19), due to the lack of the sequential consensus with any members of the known bacterial kinase families (bacterial tyrosine kinases, BY-kinases (6)) or eukaryotic-like protein kinases. PtkA is considered as the cognate kinase of MptpA and was proposed to modulate its activity by tyrosine-specific phosphorylation (17). Biochemical analysis annotates PtkA as a member of the haloacid dehydrogenase-like hydrolase (HAD) superfamily, due to the presence of the active site motif D 85 XD, which is essential for the catalytic and autophosphorylation activity (18). Furthermore, PtkA acts as substrate for Mtb endogenous eukaryotic-like STPKs (20), which might represent an additional level of regulation of PtkA activity. MptpA and PtkA, even if they do not represent typical on/off switches of signal transduction (19), are part of the Mtb regulatory system and therefore important target candidates for anti-TB therapy. Moreover, recent studies show the essential role of PtkA for Mtb growth in macrophages, establishing a central role for the PtkA-MptpA operon in Mtb pathogenesis (21). In this context, the lack of structural information of PtkA hampers a general understanding of the interactions between MptpA and PtkA, impairing the development of strategies for rational design of TB inhibitors.
Here, we present an extensive study on the structure and dynamics of the WT PtkA using NMR spectroscopy, describing the intrinsic structural architecture of PtkA at atomic resolution. We characterize the two domains that build up the PtkA structure, an N-terminal intrinsically disordered domain (IDD PtkA ) and C-terminal kinase core domain (KCD PtkA ). Additionally, we solved the three-dimensional NMR structure of KCD PtkA . Further, using a broad spectrum of NMR techniques, we characterize inter-and intramolecular interactions of PtkA, giving important insights into the key regulatory catalytic processes of tyrosine phosphorylation-dephosphorylation mediated by PtkA-MptpA.

Structural study of PtkA
The sequence-based analysis of PtkA (30.6 kDa, 291 amino acids) predicts a high disorder tendency for the N-terminal part comprising 80 amino acids (Fig. S1). Our NMR data confirm the prediction, showing that PtkA consists of a well-folded KCD and an unstructured IDD. The 2D 1 H, 15 N TROSY spectrum of the full-length PtkA shows in total 231 of 270 expected amide resonances (Fig. 1A). Two subsets of signals can be clearly observed: (i) a set of well-dispersed peaks, typical for a well-folded structure, and (ii) another set of signals with high intensity, which are clustered in the center of the spectrum ( 1 H: 7.6 -8.6 ppm), indicative of unstructured regions within the full-length PtkA. The 2D 1 H, 15 N HSQC spectrum of PtkA shows only minor changes of the chemical shift in the presence of ATP (Fig. S8). Unfortunately, incubation of PtkA

Structural studies of PtkA by NMR
with ATP for 24 h leads to precipitation of the protein, which excludes phosphorylated PtkA to be used for long-term NMR experiments.

Sequence-specific NMR resonance assignment of full-length PtkA
The assignment of the 1 H, 15 N, and 13 C backbone resonances of the full-length PtkA was obtained using a combination of standard triple-resonance NMR experiments on triple-labeled ( 2 H, 15 N, 13 C) and/or double-labeled ( 15 N, 13 C) PtkA samples. We observed that the number of visible cross-peaks increases to 20% in the presence of Mg 2ϩ , leading to better quality of the NMR spectra. Therefore, all assignment experiments were performed in the presence of 10 mM MgCl 2 . In addition, the selectively 15 N-Val-and 15 N-Phe-labeled PtkA samples were prepared to confirm ambiguous assignments. We were able to assign 93.5% of the observable backbone amides (216 of 231 detected amide resonance in the 2D 1 H, 15 N TROSY spectrum), which is 80% of the expected amide cross-peaks. Most of the missing chemical shift assignments were restricted to three regions: Thr 25 -Ser 29 , Thr 210 -Asp 219 , and Val 257 -His 278 (Fig.  1C), presumably due to intermediate exchange on the NMR time scale. To complete and verify the assignment of the fulllength PtkA, we undertook a protein construct optimization. The assignment of the full-length PtkA was thus aided by using two PtkA truncation mutants (Fig. 1B) containing either the N-terminal 81 amino acids (Met 1 -Pro 81 ) representing the intrinsically disordered domain (IDD PtkA , 8.5 kDa) or the C-terminal (Val 75 -Val 291 ) region representing the kinase core domain (KCD PtkA , 23 kDa). The structural properties of both constructs were investigated by NMR spectroscopy.
Backbone assignment of KCD PtkA -The 2D 1 H, 15 N HSQC spectrum of KCD PtkA revealed a substantial reduction of the signal overlap, and the remaining 172 well-resolved backbone amide signals fit almost perfectly with those observed in the spectrum of the full-length PtkA (with the exception of those residues that are located close the truncation site: Gly 76 , Glu 77 , and Ser 78 ), indicating proper folding of the KCD PtkA . Using the KCD PtkA in separation, we were able to obtain 96.5% of the backbone assignment (166 of 172 observable amide crosspeaks; Fig. S2A), which is 76.9% of the expected amide crosspeaks. Hence, the KCD PtkA was used for the three-dimensional NMR-based structure determination.
Backbone assignment of IDD PtkA -The presence of intrinsically disordered domain hampers the unambiguous assignment in the overcrowded spectral region of the full-length PtkA. The 2D 1 H, 15 N HSQC spectrum of IDD PtkA at pH 7.5 revealed poor signal dispersion, similar to the observation made in the spectrum of the full-length PtkA. To facilitate the backbone assignment of the IDD PtkA , we investigated the IDD PtkA also at low pH. In the 2D 1 H, 15 N HSQC spectrum of IDD PtkA measured at pH 2.0, we could detect 68 well-resolved signals and were able to assign all of the amide backbone using a set of 3D NMR experiments (HNCACB, (H)N(CA)NNH, HNHA, and HNCO) (Fig. S2C). A pH titration series of 2D 1 H, 15 N HSQC spectra were recorded, ranging from pH 2.0 to 7.5 with a 1.5-pH unit increment, to enable the backbone assignment of the 2D 1 H, 15 N HSQC spectrum of IDD PtkA at pH 7.5 (Fig. S2B). The detailed characterization of the IDD PtkA was recently published by our group (22).
The resonance assignment from the KCD PtkA (residues 76 -291) and IDD PtkA (residues 1-81) was used to validate the assignment of the full-length PtkA (residues 1-291). We were able to assign 69% of the IDD (47 signals of 68 expected) and 84% of the KCD (169 of 202 expected) backbone amide signals in the full-length PtkA at pH 7.5 (Fig. 1A). The backbone assignment of PtkA has been deposited in the Biological Magnetic Resonance Data Bank (BMRB) with entry number 34204.

3D structure determination of the KCD PtkA
The NMR solution structure of the PtkA kinase core domain was determined (Protein Data Bank entry 6F2X). Due to the limited completeness of the resonance assignment of residues located in or near to the center of the protein, the conventional structure calculation with fully automated NOESY cross-peak assignment by CYANA was unsuccessful. Missing assignments within and around two central regions (e.g. Thr 210 -Asp 219 and Val 257 -His 278 ) indicate dynamics or exchange on an intermediate NMR time scale for this part of the molecule, complicating the structural characterization of the KCD PtkA . Both regions are particularly interesting, as they contain residues involved in the catalytic regulation of PtkA. To be able to elucidate the three-dimensional structure of PtkA, homology models were used to assist the NOESY cross-peak assignment by CYANA, thereby just avoiding NOE misassignments between atoms that are actually not close to each other (assigned resonances that coincide with those of the undefined region), which would otherwise disrupt the structure and cause the convergence to fail. The subsequent structure refinement was done without the additional global distance restraints (see "Experimental procedures" for more details). This allowed us to characterize the structural propensities of the PtkA core domain (Fig. 2). The regions with missing assignments (Thr 210 -Val 216 and Tyr 262 -Val 275 , indicated in gray) are undefined due to the lack of restraints (thus merely force fieldbased), and their spread does not necessarily indicate dynamics. Six parallel ␤-sheets located in lobe 1 (Fig. 2B) of the KCD PtkA build up the hydrophobic core of the KCD PtkA . Lobe 1 is the essential part of the PtkA-containing catalytic loop (D 85 LD motif), the lysine residues essential for ATP binding (Lys 184 , Lys 217 , and Lys 270 ), and the autophosphorylation site (Tyr 262 ). Lobe 2 of the core domain consists of three long ␣-helices (designated ␣1-␣3) and a short ␣-helical section that is located between ␣-helix ␣1 and ␣-helix ␣2. Lobe 2 contains two additional tyrosine residues (Tyr 146 and Tyr 150 ) with an unknown role. However, the description of the catalytic site of PtkA by NMR spectroscopy in any detail is difficult because only a subset of the residues involved in the catalysis is amenable for interpretation, largely due to missing amide cross-peaks at nativelike pH (pH 7.5). However, mapping of the position of residues involved in the catalytic regulation of PtkA (18) onto the structure reveals that all of the three lysine residues (Lys 184 , Lys 217 , and Lys 270 ), the tyrosines (Tyr 146 , Tyr 150 , and Tyr 262 ), and the Asp 85 located in the conserved DXD motif are surrounding the autophosphorylation site (Fig. S3A). In addition, primary sequence alignment of PtkA and two putative phosphatases Structural studies of PtkA by NMR (from Clostridium difficile and phosphatase from Clostridium acetobutylicum) revealed no sequence analogy for the N-terminal IDD, whereas four conserved HAD signature motifs were found to be conserved within the C-terminal domain (Fig. S3B).

Study of PtkA dynamics: Heteronuclear 15 N relaxation studies
PtkA dynamics were studied using standard heteronuclear 15 N relaxation experiments. The experiments were performed on the full-length PtkA as well as on the KCD PtkA and the IDD PtkA to investigate whether the domains influence the dynamics of each other. The experimentally obtained relaxation rates for 15 N-T 1 , 15 N-T 2 , and 15 N{ 1 H}-heteronuclear NOE were used to determine the Lipari-Szabo order parameter, S 2 (Fig. 3A). The relaxation study of the internal motions in the full-length PtkA shows a rigid KCD and a higher flexibility of the N-terminal domain (Met 1 -Glu 77 ; indicated by an increase of the R 1 rate, a decreased R 2 rate, and a lower 15 N{ 1 H}heteronuclear NOE value). The relaxation data obtained for the KCD PtkA construct are highly similar to those for the full-length PtkA, which indicates that there is no major effect on the local backbone dynamics of the KCD in the absence of the IDD. For the IDD PtkA , we determine 15 N-R 1 relaxation rates of 1.78 Ϯ 0.13 s Ϫ1 and 15 N-R 2 rates of 3.44 Ϯ 0.33 s Ϫ1 . The relaxation rates thus obtained for the IDD PtkA and the full-length PtkA are comparable and confirm high flexibility and disordered state at native-like conditions for the N-terminal domain of PtkA.

Hydrogen-deuterium exchange studies
Hydrogen-deuterium exchange studies were used to study the solvent accessibility of the amide protons of PtkA in the presence and absence of the IDD. The exchange rates (k ex ) of the backbone amides of the full-length PtkA and the KCD PtkA were determined (Fig. 3B). Fifteen minutes after reconstitution of the freeze-dried full-length PtkA sample into D 2 O, only 33% of the backbone amide cross-peaks were observed compared with the signals in H 2 O. All amide signals corresponding to the residues of IDD disappeared, which is typical for very flexible dynamic regions and thus well in agreement with the results from our heteronuclear relaxation studies. In contrast, 10% of the amide signals of residues located in the KCD were still visible even after 19 h. These very slowly exchanging backbone amide protons correspond to the residues Val 82 -Asp 85 , Leu 167 -Arg 172 , Val 176 -Thr 182 , Leu 193 -Phe 196 , Ala 206 , Leu 225 , Val 236 -Asp 240 , and Leu 285 -Val 291 , suggesting involvement of those residues in the hydrogen bond network of highly protected secondary structure elements in the KCD. Furthermore, the correlation between the secondary structure elements and hydrogen-deuterium exchange rate reveals that the KCD is inherently stable and rigid. At the same time, different dynamic properties of the ␣-helices in the KCD were observed. Slow exchanging amide protons from residues in the loop region between ␤-sheet ␤1 and ␣-helix ␣1 and residues in the ␣-helix  Figure 3. A, PtkA dynamics before and after truncation of one of the domains. Top, cylinders, ␣-helices; arrows, ␤-sheet. Shown is a schematic representation of an NMR chemical shift-based TALOS ϩ prediction of the secondary structure elements of KCD PtkA . Shown is a plot of the determined order parameter (S 2 ) for the full-length PtkA and KCD PtkA construct as a function of residue number (S 2 was generated using TENSOR2 software (29) based on the three experimentally measured relaxation parameters: T 1 , T 2 , and 15 N heteronuclear NOE). Shown is a plot of the R 1 and R 2 relaxation rates and 15 N{ 1 H}-heteronuclear NOE as a function of residue number of the full-length PtkA (black) and KCD PtkA (red) determined at 700 MHz and IDD PtkA (blue) measured at 600 MHz, 298 K, in 50 mM HEPES-NaOH buffer (pH 7.5) containing 300 M protein, 300 mM NaCl, 10 mM DTT, 10 mM MgCl 2 , 10% D 2 O/90% H 2 O. B, hydrogen-deuterium exchange rate (K ex , s Ϫ1 ) of full-length PtkA (black) and KCD PtkA (red) as a function of residue number. The missing exchange rates could not be determined due to rapidly exchanging amide hydrogens (directly after solvation in D 2 O), for unassigned or overlapping cross-peaks, and for proline residues. C, determined temperature coefficient (ppb/K) of full-length PtkA (black), KCD PtkA (red), and IDD PtkA (blue) as a function of residue number. The error estimate is derived from the Gaussian noise in the peak intensity. ␣ 4 -6 suggest potential exposure to the hydrophobic part of the protein, whereas ␣-helix ␣ 1-3 and ␣ 6 -7 fluctuate in solution. Moreover, for PtkA without the IDD, slight differences in the backbone amide hydrogen exchange behaviors were observed after 3 h of reconstitution in D 2 O (Fig. S4). Signals corresponding to the residues Val 144 and Ala 152 , located in the ␣-helix (␣ 2 ) containing Tyr 146 and Tyr 150 ; Phe 161 , located in the ␣ 2 -␣ 3 loop; and Val 256 , located near the Tyr 262 , were missing in the fulllength PtkA but still detectable in the spectrum of KCD PtkA . We speculate that these minor differences observed may play an important role during the autocatalytic regulation of PtkA, involving tyrosine phosphorylation (7). Hence, the dynamic nature of the intrinsically disordered N-terminal domain may play a role in modulating the behavior and arrangement of the important helices involved in catalysis.

Intramolecular hydrogen bonding
A temperature series was performed to study intramolecular hydrogen bonding based on the amide proton temperature dependence. We determined the temperature coefficient (Tcoeff., in ppb/K; Fig. 3C) from the amide chemical shift values by acquiring a series of 2D 1 H, 15 N HSQC spectra measured at different temperatures, from 301 to 283 K with 3 K increments (Fig. S5). We analyzed the full-length PtkA as well as the KCD PtkA and IDD PtkA , to examine the involvement of the amide protons in the formation of rigid or transient hydrogenbonded structure elements. Amide protons with a temperature coefficient more negative than Ϫ4.5 ppb/K (indicated by the dotted line in Fig. 3C) are generally not hydrogen-bonded (23). Many values obtained for amide protons of residues located in the KCD are larger than Ϫ4.5 ppb/K, which clearly suggests their involvement in the hydrogen bond formation, which is in agreement with the presence of abundant secondary structural elements in this domain. On the other hand, for the IDD, nearly all values are below Ϫ4.5 ppb/K, indicating that the residues in this domain are generally not hydrogen-bonded. Only Asn 48 (Ϫ3.7 ppb/K), Gly 49 (Ϫ3.6 ppb/K), and Asn 60 (Ϫ4.0 ppb/K), which show a temperature coefficient larger than Ϫ4.5 ppb/K, were observed, indicating a potential involvement of those residues in transient hydrogen bonding.

PtkA intramolecular IDD-KCD interaction
The intramolecular domain-domain interactions were examined by comparing the chemical shifts observed in 2D 1 H, 15 N HSQC spectra of the full-length PtkA, IDD PtkA , and KCD PtkA (Fig. S6, A and B). The spectra of the full-length PtkA and the KCD PtkA indicate that the folding of the KCD is not affected by truncation of the IDD. The comparison of the amide chemical shift of the IDD in the full-length PtkA and the IDD PtkA spectra shows the same chemical shift for most of the cross-peaks, except for the residues located proximate to the truncation site as well for Asn 48 (Fig. S6A), suggesting weak domain-domain interactions. To study whether such interactions actually occur, we performed NMR titration experiments between 15 N-labeled KCD PtkA and unlabeled IDD PtkA (up to molar ratio 1:2) and vice versa (up to molar ratio 1:3) at 298 K and pH 7.5 in the presence and absence of ATP (10 mM). The 2D 1 H, 15 N HSQC spectra recorded during the titration series revealed no significant CSPs (Fig. S6, C and D). Based on these results, we assume that interactions between the IDD PtkA and KCD PtkA in trans (IDD and KCD are separated by mutational deletion), if present, are transient and rather weak. Nevertheless, information based on the amide chemical shift is primarily limited to the protein backbone and may thus be insufficient to detect weak interactions. Paramagnetic relaxation enhancement (PRE) NMR spectroscopy has been efficiently used to study weak and transient interactions. To detect these potential weak interdomain interactions of PtkA, we used the site-directed spin label (SDSL) approach. With SDSL, we analyzed long-range PtkA domain-domain interactions in cis (native state in full-length PtkA). The nitroxide radical, 1-oxy-2,2,5,5tetramethyl-D-pyrroline-3-methyl-methane-thiosulfonate (MTSL), was covalently attached to the conserved cysteine (Cys 61 ) residue located on the IDD. Further, to investigate the presence of residual structure in the IDD (Fig. S1), two single-cysteine PtkA mutants were designed, including mutations on alanine 10 (PtkA/C61A/A10C) and serine 41 (PtkA/C61A/A10C). The point mutations were introduced far away from the potential predicted secondary structure element, such that it did not disturb the fold (A10C), and a second mutant (S41C) was introduced within the folded region to interrupt the formation of the secondary structural element. After spin labeling with MTSL, 2D 1 H, 15 N HSQC spectra of the paramagnetic and diamagnetic protein (after the addition of ascorbic acid) were acquired. In all spectra of the protein containing a paramagnetic center, PREinduced line-broadening effects (Fig. 4A) were observed. The intensity ratio analysis of paramagnetic and diamagnetic (I para / I dia ) PtkA resolved the most affected regions (Fig. 4B). For all three of the investigated PtkA constructs, the same regions in KCD were affected (I para /I dia Յ 0.4, Glu 114 -Gly 134 , Asp 162 , Thr 188 -Ile 192 , Ile 205 -Gly 212 , Leu 231 -Met 237 , Trp 260 , and Ile 282 ), suggesting a preferential position of the IDD relative to the core domain and the possibility of transient long-range domain-domain interactions. To visualize the IDD-KCD interaction, a set of experimental restraints from the spin-labeled complexes was used for the calculation of the model shown in Fig. 4C. Analysis of this model shows the presence of multiple orientations for the IDD within the IDD-KCD complex, suggesting a dynamic nature for this interaction. Further, the ensemble of similar orientation indicates that the IDD transiently interacts with the KCD and masks the ATP-binding site (Fig. 4C, right).

PtkA autophosphorylation and regulation of its catalytic activity
The catalytic activity of PtkA is regulated via autophosphorylation of tyrosine residues in the presence of phosphate donor ATP or GTP (18). For the transfer of the phosphoryl group (PO 3 2Ϫ ) from ATP to the target protein and the formation of the physiological substrate XATP 1Ϫ , the kinase requires divalent cations (e.g. X ϭ Mn 2ϩ , Mg 2ϩ ). Three tyrosine residues (Tyr 146 , Tyr 150 , and Tyr 262 ) are present in the sequence of PtkA and represent potential sites for autophosphorylation. In particular, Tyr 262 was considered as the target for phosphorylation during autocatalysis (18). PtkA itself represents a target protein for phosphorylation by the Mtb endogenous STPK (20). We inves-Structural studies of PtkA by NMR tigated (i) the autophosphorylation of PtkA, (ii) the effect of Mg 2ϩ , (iii) interactions between PtkA in the nonphosphorylated and phosphorylated state with MptpA, (iv) dephosphorylation of PtkA by MptpA, and (v) PtkA phosphorylation by serine/threonine kinase (Fig. S11) (22).
PtkA autophosphorylation-The autophosphorylation activity of full-length PtkA was previously measured and confirmed using a luciferase assay (16). Furthermore, the autophosphorylation site in PtkA resides on the KCD and not on the IDD. To test the influence of the IDD on the autophosphorylation of PtkA, we performed a luminescence assay using a KCD PtkA construct alone and in the presence of the IDD PtkA and normalized the activity to that of the PtkA, which was set to 100% ( Fig. 5C  and Fig. S12). The results thus obtained suggest that the KCD PtkA is 5 times more active than the full-length PtkA. Activity of the KCD PtkA measured together with isolated IDD (IDD PtkA ) shows a decrease compared with the KCD PtkA alone, suggesting that the IDD has an inhibitory effect on the kinase activity. The IDD in PtkA presents several potential PTM (phosphorylation) sites. Surprisingly, the activity of the PtkA in which the IDD was phosphorylated (PtkA*-P) showed 4 times more activity than that of its nonphosphorylated form (PtkA*) (Fig. 5C). These results strongly suggest that the IDD transiently binds to the KCD and masks the substrate-binding site and inhibits the kinase activity. Using 1D 31 P NMR spectroscopy, we monitored the autophosphorylation reaction of PtkA in the presence and absence of the IDD after overnight incubation of PtkA (100 M) with ATP (10 mM) at pH 7.5, 25°C (Fig.  5D). In both cases, an additional signal at Ϫ0.98 ppm, indicative of the phosphorylated state of PtkA, was observed. MALDI-MS confirmed that PtkA was phosphorylated one time (Fig. S10).
PtkA interactions with Mg 2ϩ -To explore the effect of Mg 2ϩ on the structural propensities of PtkA, we performed NMR studies in the presence and absence of MgCl 2 . The addition of MgCl 2 (10 mM) into NMR buffer greatly improves the quality of the 2D 1 H, 15 N HSQC spectrum of PtkA, resulting in 20% more detectable amide cross-peaks. Comparison of the chemical shift of the full-length PtkA obtained in the presence and absence of MgCl 2 allowed us to delineate strong Mg 2ϩ -induced CSPs. The most affected resonances come from residues of the KCD. Additional experiments using KCD PtkA showed a similar effect of Mg 2ϩ (Fig. S7).

Structural studies of PtkA by NMR
PtkA-MptpA interface-In addition to previously reported MptpA-PtkA interaction studies (16), for which the interaction was only mapped on the MptpA protein, we performed a set of NMR titrations for mapping the interaction interface on PtkA. For these studies, 15 N-labeled full-length PtkA as well as 15 N-labeled KCD PtkA were titrated with unlabeled MptpA (up to molar ratio 1:4) in the presence or absence of 5 mM ATP. Similar to the interaction studies from the MptpA, we obtained a number of small but evident CSPs, which were consistent for both of the constructs (PtkA and KCD PtkA ) (Fig. S8, A and B).
The largest number of chemical shift changes were observed for the residues surrounding the catalytic loop (D 85 LD), located between ␤-sheet (␤1) and ␣-helix (␣1). Additionally, residues located in the ␣-helix (␣5) and ␤-sheet (␤3) that envelops the autophosphorylation site are also affected, indicating their involvement in the binding interfaces with MptpA. Furthermore, mapping of the CSPs onto the structure of PtkA shows slight differences of the affected KCD area in the presence and absence of IDD. The ␣2-␣3 loop located in lobe 2 of KCD seems to be more involved in the interactions in the absence of IDD.

Structural studies of PtkA by NMR
Interestingly, the ␣3-helix, which contains the two tyrosine residues Tyr 146 and Tyr 150 also showed similar differences in the hydrogen-deuterium exchange experiments measured in the presence and absence of IDD.
Dephosphorylation of PtkA by MptpA-To study the dephosphorylation of PtkA by MptpA, we synthesized peptides derived from PtkA, representing the autophosphorylation site in (i) nonphosphorylated (Asp 264 -Lys 270 /Tyr 262 ) and (ii) phosphorylated Tyr 262 (Asp 264 -Lys 270 /pTyr 262 ) state. 1D 31 P NMR of the PtkA (Asp 264 -Lys 270 /pTyr 262 ) peptide shows a signal at Ϫ0.17 ppm, indicating phosphorylation of Tyr 262 (Fig. 5E). We monitored this signal of the PtkA peptide (Asp 264 -Lys 270 / pTyr 262 ) both upon the addition of and in the absence of the phosphatase MptpA (Fig. S9). Dephosphorylation of PtkA peptide (Asp 264 -Lys 270 /pTyr 262 ) is indicated by the disappearance of the signal in the 1D 31 P NMR spectra at Ϫ0.17 ppm and the appearance of one additional signal at 2.29 ppm corresponding to the P i . The aromatic signals (H␦ and H⑀) of Tyr 262 reappear at a different chemical shift upon dephosphorylation by MptpA. In addition, the 2D 1 H, 13

Discussion
Noncanonical bacterial kinases possessing tyrosine kinase activity play an important role in the cellular regulation of bacteria. Those kinases possibly belong to various protein families with multiple functions, as there is no sequential consensus between them. Besides PtkA from M. tuberculosis (18), other kinases, such as DivL of Caulobactercerescentus (24), WaaP of Pseudomonas aeruginosa (25), or PutA of Salmonella typhimurium (26), were found and termed as odd PTKs (19). How widespread such tyrosine kinases are in prokaryotes is still unclear, but the existence of atypical tyrosine kinases within one species suggests the existence of additional members of the same subfamily. The presence of homologs in the entire operon (similar to that found for the protein-tyrosine kinase PtkA of Mtb) in multiple Actinomycetes species like Rhodococcus, Corynebacterium, Gordonia, or Amycolicicoccus suggests that these proteins, even with limited homology, are protein-tyrosine kinases. Furthermore, as many as 924 unique PtkA orthologs were identified in all of the domains of life (19).
The tyrosine phosphorylation and dephosphorylation represent a significant part of the survival regulation of Mtb (27). The interaction of Mtb protein-tyrosine phosphatase MptpA with the host defense machinery plays an important role in the mycobacterial virulence (17). Moreover, PtkA was recently shown to play a central role in promoting the growth of Mtb in macrophages (21), further underlining the significance of the interaction studies between MptpA and its cognate kinase PtkA and its role in Mtb regulation. Both enzymes represent potential candidates for rational drug design, where inactivation of either one of the partners or their interaction may inhibit the pathogen survival. Furthermore, the PtkA-MptpA operon conserved among numerous Actinobacteria species with lim-ited identity for PtkA (39 -76%) is in contrast to the Mtb complex, where this operon is highly conserved, with 99.7-100% protein identity (19). This additionally indicates the important role of this protein in the Mtb physiology.
We characterized the structure of PtkA and investigated its interactions with MptpA in detail using NMR spectroscopy. The structural architecture of PtkA deviates from other bacterial kinases, as it contains an N-terminal intrinsically disordered domain, which is linked to the well-folded kinase core domain. The KCD is the catalytic center of PtkA containing (i) a catalytic loop (D 85 LD motif) located between the ␤-sheet (␤1) and ␣-helix (␣1) secondary structure and (ii) the autophosphorylation site (Tyr 262 ) located near the protein hydrophobic core. The PtkA conformation is stabilized by binding of a divalent cation Mg 2ϩ , which may regulate the substrate binding. Moreover, MptpA binds to the KCD of PtkA and dephosphorylates the kinase. The dynamic properties of both domains are very different and suggest the possibility of an interdomain controlled regulation of PtkA activity. The IDD is unstructured and highly dynamic, allowing for a "fly-casting-like" mechanism of transient interactions with the rigid KCD. This interaction thereby regulates the accessibility of the KCD active site (which is involved in autophosphorylation). The decrease of the PtkA activity in the presence of IDD suggests an inhibitory effect of the disordered domain during the autophosphorylation. Spin label studies indicated the existence of residual long-range transient interactions of the IDD with the KCD. Based on those results, we made a model of the IDD-KCD structure, which shows that the preferred position of the IDD is located near the catalytic site. Our studies show that the presence of the IDD does not affect the fold of the KCD and has only minor effects on the solvent accessibility of the KCD. The changes in dynamics are restricted to the upper lobe of the KCD (␣-helix (␣3)), which harbors two tyrosines whose role and phosphorylation status are unclear. The kinase activity was observed in the presence as well as in the absence of IDD, which is in agreement with the results obtained in 1D 31 P NMR spectra, where in both cases the phosphorylation of PtkA was detected. However, our activity studies show that the IDD has an inhibitory effect on the PtkA enzymatic activity. This suggests that the IDD masks the region of Tyr 262 , thereby restricting the accessibility of this key residue involved in the autophosphorylation of PtkA. Based on the above observations, we propose two conformational states for PtkA: (i) open state, where the IDD is away from the autophosphorylation site, increasing the accessibility of Tyr 262 for the phosphorylation, and (ii) closed state, where the Tyr 262 is masked by the IDD. Multisite phosphorylation of IDD releases the IDD from the KCD, thereby uncovering Tyr 262 and consequently increasing the PtkA activity. Phosphorylation of PtkA by serine/threonine kinase induces conformational changes of IDD, which abolish IDD movement and promote the open state of PtkA, where the Tyr 262 is accessible to the phosphorylation. However, the exact catalytic mechanism behind the modulation of the PtkA activity can be a subject for further investigation.
Based on our findings, we propose the following model to explain the potential function for this unusual IDD in PtkA, in the context of overall regulation of Mtb virulence (Fig. 6).

Structural studies of PtkA by NMR
MptpA is secreted from the mycobacterial cytosol to the cytosol of infected macrophages by an unknown translocation mechanism. During the phagosome-macrophage fusion, MptpA migrates into the macrophage cytosol. The phosphorylation state of MptpA at this point is unclear; however, we know that activated PtkA catalyzes the phosphorylation of MptpA, thereby increasing its activity (17) which possibly plays a key role in the translocation of MptpA (Fig. 6, left inset) to the cytosol of infected macrophages. Moreover, multisite phosphorylation of IDD of PtkA shifts the balance between the open and closed states of the IDD toward a more open state. As a consequence of this PTM of the IDD, PtkA might now possess enhanced activity toward phosphorylating MptpA and might also become anchored to the Mtb cell wall (Fig. 6, right). Phosphorylation-driven membrane anchoring has been observed previously for protein kinase C (28). In summary, the structural and functional knowledge of PtkA gained in this study is crucial for understanding the MptpA-PtkA interactions, catalytic mechanism, and in general the role of the kinase-phosphatase regulatory system in Mtb virulence.

Cloning, expression, and purification of PtkA
The pET151/D-TOPO plasmid encoding WT PtkA (Rv2232, Met 1 -Val 291 ) was transformed into Escherichia coli BL21 (DE3) cells for expression. The PtkA subdomains IDD PtkA (Met 1 -Leu 81 ) and KCD PtkA (Gly 76 -Val 291 ) were generated by PCR. WT PtkA DNA was used as a template for DNA replication. The PCR product was amplified into the pRS-45 vector using NedI and BamHI as restriction sites. The efficiency of the insertion was confirmed via nucleotide sequencing. The PtkA DNA plasmid generated via PCR was transformed into E. coli BL21 (DE3) for protein expression. The expression of recombinant PtkA was performed in LB medium for unlabeled protein and in M9 medium for 15 N/ 13 C uniformly labeled PtkA. Cell cultures (supplemented with 1 mM ampicillin) were grown at 37°C with aeration (120 rpm) to A 600 nm ϳ0.7, incubated at 0°C for 15 min, and induced using 1 mM isopropyl 1-thio-␤-D-galactopyranoside for overexpression. Triple ( 2 H, 13 C, 15 N)-labeled PtkA was expressed in rich growth medium solution labeled with stable isotopes ( 2 H (Ͼ95%), 13 C, 15 N) (Silantes). After adaptation of the cells to 70 and 100% D 2 O at A 600 nm ϳ0.5, the cells were washed twice in PBS/D 2 O and resuspended in E. coli-OD 2 CDN Silantes medium at A 600 nm ϳ0.7. The protein expression was induced at A 600 nm ϳ1.0 using 1 mM isopropyl 1-thio-␤-D-galactopyranoside. After incubation overnight for unlabeled or uniformly 15 N/ 13 C-labeled PtkA and 12 h for 2 H, 13 C, 15 N-labeled PtkA at 16°C with aeration (120 rpm), the cells were centrifuged (4,000 ϫ g, 45 min, 4°C), and the cell pellet was resuspended in lysis buffer (300 mM NaCl, 50 mM Tris-HCl, pH 8.0, 10 mM 2-mercaptoethanol), supplemented with complete protease inhibitor (1 tablet/100 ml, EDTA-free,

Structural studies of PtkA by NMR
Roche Applied Science). The cells were disrupted for 15 min using M-110P microfluidizer (15,000 p.s.i.). The supernatant was separated from cell residues by centrifugation (18,000 ϫ g, 45 min, 4°C). The soluble protein fractions were applied to a nickel-nitrilotriacetic acid Fast Flow Column (GE Healthcare), following the manufacturer's recommendations. His 6 -tag cleavage was performed via dialysis (300 mM NaCl, 50 mM Tris-HCl, pH 8.0, 10 mM 2-mercaptoethanol) overnight at 4°C using tobacco etch virus protease (1 mg/ml) and separated using the nickel-nitrilotriacetic acid Fast Flow Column. Subsequently, preparative size-exclusion chromatography was performed on a HiLoad 26/60 Superdex 75 column (GE Healthcare) in running buffer (50 mM HEPES-NaOH, pH 7.5, 300 mM NaCl, and 10 mM DTT) to increase protein purity. The eluted fractions were analyzed by SDS-PAGE. Fractions containing pure protein were pooled and stored at Ϫ80°C or immediately used for further experimental procedures.

Peptide synthesis
The solid-phase peptide synthesis was carried out by standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. The C-terminal residue was manually attached to a 2-chlorotrityl chloride resin. Peptides were purified by reversed-phase HPLC and characterized using electrospray ionization MS and analytical HPLC.

Luciferase assay
The autophosphorylation activity of the protein was determined using a Kinase-Glo luminescent kinase assay (Promega). 20 M PtkA was added to the assay buffer (300 mM NaCl, 50 mM Tris-HCl, pH 8.0, 10 mM DTT) containing different amounts of ATP (1,2.5,5,7.5,10 M;Fermentas) in the presence of 10 mM MgCl 2 . After overnight incubation at 25°C, the assay mix (50 l) was applied on to a 96-well plate (white; E&K Scientific EK-25075), and the luciferase mixture (50 l) was added. After incubation for 2 h, the luminescence was measured using a Veritas TM microplate luminometer.

Autophosphorylation reaction
The autophosphorylation reaction of PtkA was performed using unlabeled 100 M protein and 10 mM ATP in 50 mM HEPES-NaOH, pH 7.5, buffer containing 300 mM NaCl, 10 mM DTT, and 10 mM MgCl 2 . After incubation overnight at 25°C, MALDI-MS as well as 1D 31 P NMR were measured.

Resonance assignment experiments
For the backbone assignment of PtkA, a set of 3D tripleresonance experiments were collected, including HNHA, HNCO, HNCA, HN(CO)CA, HNCACB, and HN(CO)CACB experiments on double-and/or triple-labeled protein. The side-chain assignment of the aliphatic resonances was obtained using HCCCONH and CCCONH experiments. 2D HBCB-CGCDHD and HBCBCGCDCEHE spectra and an aromatic 13 C-filtered TOCSY spectrum were used for the assignment of the aromatic side-chain resonances. For assignment and NOEbased distance restraints, 3D 1 H, 1 H, 15 N NOESY-HSQC (mixing time, 120 ms) and aromatic and aliphatic 3D 1 H, 1 H, 13 C NOESY-HSQC (mixing time, 75 ms and 120 ms) in H 2 O were collected. 15 N relaxation experiments were performed at 298 K on a 700-MHz spectrometer. The R 1 longitudinal 15 N relaxation rates were obtained from a series of experiments acquired with varying delays of 100, 200, 400, 600, 800, 1200, 1600, 2000, 2400, and 3000 ms. R 2 transverse 15 N relaxation rates were determined from a series of spectra using the following delays: 16 H}-heteronuclear NOE values were obtained from the analysis of peak intensity ratio (I NOE /I NONOE ) from the recorded spectra with and without saturation of amide protons. The order parameter, S 2 , was determined using TENSOR2 (29).

Hydrogen-deuterium exchange
For the hydrogen-deuterium exchange experiment, 150 l (200 M) of protein in NMR buffer was lyophilized and dissolved in the equivalent volume of deuterium oxide (D 2 O). A series of 2D 1 H, 15 N HSQC spectra were recorded immediately after the addition of D 2 O, for a duration of 15 min within the first hour and subsequently for 30 min, for a total duration of 19 h. As a reference for the evaluation of spectra, a 2D 1 H, 15 N HSQC spectrum of the protein (200 M) in NMR buffer (with 90% H 2 O/10% D 2 O) was measured before. The hydrogendeuterium exchange rate for each residue was determined by fitting the amide peak intensities to exponential decays over the time for each residue (I(t) ϭ I(0)⅐exp(Ϫk ex ⅐t)), where I(t) and I(0) are the intensities at the given time t and t ϭ 0, and k ex is the rate constant of the hydrogen-deuterium exchange reaction.

Temperature series
The 2D 1 H, 15 N HSQC spectra were recorded at 3 K intervals ranging from 283 to 301 K on a 700-MHz spectrometer.

Structural studies of PtkA by NMR Paramagnetic spin labeling
Single-cysteine mutations of PtkA (which contains one WT cysteine) were engineered using the QuikChange site-directed mutagenesis protocol (Invitrogen). The WT cysteine was mutated to alanine (C61A). This point-mutated DNA was used as a template for preparing single-cysteine mutations at Ala 10 (C61A/A10C) and Ser 41 (C61A/S41C). All mutations were confirmed by DNA sequencing of the entire ORF (Eurofins). Protein mutants were expressed in competent E. coli cells (BL21 (DE3), New England Biolabs), grown in M9 minimal medium enriched with 15 N-ammonium chloride, and purified as described above. The SDSL was carried out using the MTSL spin label. Protein was dissolved in the DTT-free, 50 mM HEPES/NaOH (pH 7.5) labeling buffer, containing 300 mM NaCl, 10 mM MgCl 2 and concentrated to 500 M using Vivaspin concentrators (molecular weight cut-off of 10,000; GE Healthcare). For spin labeling, a 10-fold molar access of MTSL over cysteine was added. MTSL stock solution was prepared by dissolving 10 mg of MTSL in 50 l of DMSO. After the addition of MTSL to the protein solution, the reaction mixture was incubated in the dark overnight at 25°C (WT) and 4°C (C61A/A10C and C61A/S41C). To remove excess spin label, the labeling buffer was exchanged (50 mM HEPES-NaOH, pH 7.5, 300 mM NaCl, 10 mM MgCl 2 ) using a PD-10 desalting column (Sephadex G-25 medium, GE Healthcare). The modified protein was concentrated to 300 -500 M (Vivaspin concentrators, molecular weight cut-off of 10,000) for NMR. The spin label was reduced by the addition of 5 mM ascorbate from a freshly prepared stock solution of 500 mM, to yield an ϳ10-fold molar excess of reducing agent over the spin label.

NMR data analysis
All spectra were processed using Topspin version 3.2 (Bruker Biospin) and analyzed using SPARKY version 3.114 (30).

Structure calculation
Structure calculations were performed using the software packages CYANA (31)(32)(33) and ARIA/CNS (34,35). A conventional structure calculation with fully automated NOESY crosspeak assignment failed due to the limited completeness of the resonance assignment (about 75%), especially because the missing assignments reside in or near the center of the protein.
Generally, a completeness of more than 90% (33) would suffice for a reliable automated NOESY assignment. The NOESY cross-peaks originating from unassigned regions may be misinterpreted to distances between atoms with coinciding assigned resonances, consequently disrupting the structure and causing the convergence to fail. To circumvent this problem, we made use of initial homology models to assist the initial iterative structure calculation with automated NOESY cross-peak assignment by CYANA. These structures were built by Swiss-Model (37) using a selection of homologous structures (e.g. Protein Data Bank codes 3MC1, 3SD7, 4EX6, 2AH5, 2NYV, 2YY6, 2HCF, 2HI0, 2HDO, 2HDZ, 4EEL, 3L5K, 2FDR, and 3KLZ), which have a proper range (around residues 80 -289) and sufficient sequence identity (above 20%). For the regions that are consistent with the TALOS-N secondary structure prediction (38), very ample CA-CA distance restraints (up to 20 Å) were extracted, which merely define the global shape of the molecule, thereby just avoiding NOE misassignments between atoms that are actually not close to each other. The unambiguous assigned NOEs in the 3D 1 H, 1 H, 13 C NOESY-HSQC (aliphatic and aromatic) and 3D 1 H, 1 H, 15 N NOESY-HSQC spectra were validated and inspected by using of Sparky version 3.114 (30) and subsequently used for the structure refinement (without the additional global distance restraints). The chemical shift tolerances were set to 0.015 and 0.025 ppm for the bound protons and other protons, respectively, and 0.20 ppm for the heavy atoms. In addition to NOE data, hydrogen bond distances and dihedral angle restraints for the secondary structure elements (based on TALOS-N predictions and confirmed with the Swiss-Models and initial structure calculations), as well as 1 D (HN) residual dipolar couplings and 3 J(H N ,H␣) coupling constant restraints were included in the structure calculation with CYANA (100 structures/iteration, 15,000 refinement steps). The final bundle of 20 best (lowest-energy) structures was used as input for a refinement in explicit water with CNS 1.1 (39) using the ARIA 1.2 setup and protocols (40). The 3 J(H N ,H␣) coupling constants were obtained via a 3D H N ,H␣ HMQC experiment (41,42) and directly included in the structure calculation by using the Karplus relationship. 1 D (HN) residual dipolar couplings were measured in Pf1 bacteriophages (6 mg/ml, strain LP11-92, ASLA Biotech) as alignment medium at 700 MHz and 298 K. The 1 D (HN) were extracted from IPAP-15 N, 1 H HSQC (43) spectra. Residual dipolar couplings were included in the structure calculation as direct susceptibility anisotropy restraints (SANI) and examined using the program PALES (44). The rhombicity and axial coefficients for the final alignment tensor were as follows: R ϭ 0.2 and D ϭ 4.6 with a correlation coefficient r ϭ 0.97. Moreover, 15 N relaxation data (T 1 /T 2 values) were included as diffusion anisotropy restraints in this final refinement stage. The values for the anisotropy (1.2) and rhombicity (0.3) of the rotational diffusion tensor from the final structure bundle were determined using the program Tensor 2 (29). The impact of the diffusion restraints on the structure refinement is, however, negligible, as the anisotropy of the molecule turned out to be moderate and smaller than the suggested minimal value of 1.5 (45). The overall rotational correlation time (tc ϭ 12.9 ns) is in agreement with the value predicted for a monomeric protein (tc ϭ 14.4 ns at 298 K) by the program HYDRONMR (46).