New potential eukaryotic substrates of the mycobacterial protein tyrosine phosphatase PtpA: hints of a bacterial modulation of macrophage bioenergetics state

The bacterial protein tyrosine phosphatase PtpA is a key virulence factor released by Mycobacterium tuberculosis in the cytosol of infected macrophages. So far only two unrelated macrophage components (VPS33B, GSK3α) have been identified as PtpA substrates. As tyrosine phosphatases are capable of using multiple substrates, we developed an improved methodology to pull down novel PtpA substrates from an enriched P-Y macrophage extract using the mutant PtpA D126A. This methodology reduced non-specific protein interactions allowing the identification of four novel putative PtpA substrates by MALDI-TOF-MS and nano LC-MS: three mitochondrial proteins - the trifunctional enzyme (TFP), the ATP synthase, and the sulfide quinone oxidoreductase - and the cytosolic 6-phosphofructokinase. All these proteins play a relevant role in cell energy metabolism. Using surface plasmon resonance, PtpA was found to bind immunopurified human TFP through its catalytic site since TFP-PtpA association was inhibited by a specific phosphatase inhibitor. Moreover, PtpA wt was capable of dephosphorylating immunopurified human TFP in vitro supporting that TFP may be a bona fide PtpA susbtrate. Overall, these results suggest a novel scenario where PtpA-mediated dephosphorylation may affect pathways involved in cell energy metabolism, particularly the beta oxidation of fatty acids through modulation of TFP activity and/or cell distribution.

Mtb has two PTPs, PtpA and PtpB, which are delivered into the macrophage during infection acting as key virulence factors [5][6][7][8] . PtpA and PtpB lack protein export signal sequences but both have been detected in the culture filtrates of Mtb grown in vitro 9,10 . Recent studies suggested that the SecA2 and ESX/type VII export systems are possible candidates responsible for PtpA export 4,11 . In addition, PtpA has been detected into the infected host macrophage cytoplasm by immuno-electron microscopy and Western blot analysis of the cytosolic fractions, and by the expression of neutralizing single-chain anti-PtpA antibodies 7 . The Mtb PtpA-deletion mutant strain showed reduced survival of the mycobacteria in infected human THP-1 derived macrophages, and expression of PtpAneutralizing antibodies and inhibitors simulated this effect 7,8 . Recently, Chauhan et al. generated an Mtb mutant (MtbDmms) by disrupting three virulence genes encoding PtpA, PtpB and the acid phosphatase, SapM 12 . This mutant displayed a significantly reduced ability to infect and grow inside human THP-1 macrophages. Moreover, no bacilli were recovered in spleens and lungs of guinea pigs 10 weeks following inoculation with MtbDmms, suggesting an important role of these phosphatases in the colonization of these organs 12 .
PtpA is a member of the low-molecular-weight PTP (LMW-PTP) family, which does not require metal ions for catalysis 13 . PtpA shows 37% of sequence identity and high structural similarity to its human orthologue HCPTPB. Surprisingly, the human genome encodes for 107 PTPs but only one belongs to the PtpA family which originates four protein isoforms by alternative splicing of a single transcript 14 . Two eukaryotic cytoplasmic proteins were reported as Mtb PtpA substrates: VPS33B (Vacuolar Protein Sorting 33B) which is part of the protein complex C required for membrane trafficking and fusion 7 , and the GSK3a (Glycogen Synthase Kinase 3, alfa subunit) 15 . Dephosphorylation of these macrophage components would act as a bacterial mechanism to adapt to macrophage defense response 7,16 . On one hand, dephophorylation of VPS33B by PtpA seems to exclude host vacuolar-H 1 -ATPase leading to inhibition of phagosome acidification and maturation 7,16 . Secondly, GSK3a dephosphorylation by PtpA would promote an anti-apoptotic pathway, favoring pathogen survival within host macrophage. As tyrosine phosphatases are capable of utilizing multiple protein substrates, thereby providing versatility in phospho-relay signaling networks, the search for specific phosphatase targets is still open and presents an experimental challenge.
The most commonly used biochemical tool for identifying potential PTP substrates is based on the generation of phosphatase mutants (substrate trapping mutants) that retain the ability to bind substrates but are either unable or severely impaired in carrying out substrate dephosphorylation, allowing the isolation of the PTP-substrate complex [17][18][19] . One of the most common mutants is produced by the substitution of the conserved catalytic aspartate, which assists the E-P formation and hydrolysis, by an alanine residue (D/A mutant). In the Mtb PtpA, the conserved catalytic aspartate is the Asp 126, which is not considered crucial in defining substrate specificity 20 . As reported in a kinetic study, the Mtb PtpA D126A mutant is characterized by a reduced activity (lower k cat ) compared to the wt, without substantial K m modification 21 , as also observed for the corresponding mutants of other PTPs 17,22 . This methodology has been successfully used in the identification of substrates of eukaryotic PTPs 18,19 but only a few substrates of bacterial PTPs 7,23 . The success of this methodology depends on the use of strict conditions during association and washes steps in order to avoid capturing unspecific and abundant proteins. Furthermore, it is often assumed that substrate-trapping mutants retain the structural and substrate binding properties of wt PTPs. However, significant differences may occur, leading to erroneous interpretation and thus invalidating the strategy 24 . Thus, validation of candidate substrates identified using substrate trapping is indispensable.
In this work, we attempted to improve the substrate trapping methodology to obtain novel Mtb PtpA substrates. For this aim, we firstly verified the structural and biochemical properties of the PtpA D126A mutant to ensure its adequacy for substrate trapping. Then, we prepared an extract of human macrophage-like THP-1 cells preserving phospho-tyrosine (P-Y) modifications and studied by SPR how PtpA interacted with potential substrates present in this extract. This allowed us to choose stringent experimental conditions to apply during substrate trapping steps to reduce non-specific interactions. Using this improved methodology, we successfully isolated and identified four new putative eukaryotic Mtb PtpA substrates. Three are proteins synthesized in the cytosol and then translocated to the mitochondria: (i) the alpha subunit (ECHA) of the trifunctional enzyme (TFP), an essential enzyme of the fatty acid beta oxidation pathway; (ii) the ATP synthase alpha subunit (ATPA); (iii) the sulfide quinone oxidoreductase (SQRD). The fourth protein is the 6-phosphofructokinase (K6PP) a key regulatory enzyme of the glycolysis which localizes in the cytoplasm. We advanced in the validation of one of these candidates, showing in vitro that the TFP is a bona fide substrate of PtpA.

Results
Structural and functional properties of PtpA D126A substrate trapping mutant. To achieve an adequate substrate trapping tool, the mutation introduced into PtpA should not generate substantial structural changes that could impair its functionality. Thus, in this study we generated the PtpA D126A mutant by site-directed mutagenesis and determined its structural and functional properties to assess its suitability for substrate trapping assays. We found that PtpA D126A behaved similarly to the PtpA wt through the whole expression and purification procedure, suggesting that this mutation did not significantly alter the enzyme structure ( Supplementary Fig. 1). As found for the wt enzyme, the bulk of PtpA D126A was detected in the soluble fraction following overnight induction at 15uC. Interestingly, the yield of the purified protein, estimated from three biological replicates, was higher for the mutant than for the PtpA wt (8.8 6 3.13 and 4.6 6 0.24 mg of protein per liter of bacterial culture, respectively). This difference may be associated with their respective different level in phosphatase activity (see below). In fact, expressing a full active PTP in an organism that already possesses a tyrosine kinase/ phosphatase system, as E. coli 25 , may cause a cell phosphorylation unbalance and affect the production of the recombinant protein.
With regard to the molecular mass, PtpA D126A and PtpA wt were studied by analytical SEC to determine their physical state in aqueous solution. Both phosphatases were eluted in a single major peak with an apparent molecular weight corresponding to the monomeric state (Fig. 1A).
In addition, circular dichroism studies (CD) were undertaken, to assess if the introduced mutation altered PtpA D126A secondary structure 26,27 . As shown in Figure 1B, the CD profiles of the wt and PtpA D126A mutant are basically equal, resulting in 34% a-helical structure, 19% b-strands and 47% random coils for the mutant, and in 37% a-helices, 11% b-strands and 52% random coils for the wt. These results are in agreement with the wt crystal structure (PDB ID: 1U2P) 20 . Nonetheless, PtpA D126A had a slightly reduced CD intensity in the 208 nm negative peak compared to the wt, suggesting that there is a small rearrangement in the protein conformation of this mutant. This slight change did not affect the trypsin digestion pattern of PtpA, as the same SDS-PAGE profile was observed after trypsin digestion for 1 to 3 hours (Fig. 1C). In thermal denaturation experiments the wt and mutant PtpA showed a similar profile between 25uC and 40uC, but at higher temperatures a loss of secondary structure was observed for the mutant protein indicating that the Asp 126 molecular contacts contribute to the thermal stability of PtpA wt (Fig. 1D). The kinetic properties of the PtpA D126A mutant were then examined using pNPP, a conventional artificial substrate of phosphatases. An optimal D/A substrate trapping mutant is expected to have a similar K m but a decreased value of k cat when compared to the wt enzyme 21,22 . In our case this rule was fulfilled as three independent batches of purified PtpA wt and PtpA D126A presented similar values of K m (7.4 6 3.1 mM and 5.3 6 2.3 mM, respectively) while the k cat for the PtpA D126A (0.04 sec 21 ) was 40-fold lower than that observed for the PtpA wt (1.58 sec 21 ) (Supplementary Fig. 2A). These enzymatic properties reproduced data previously reported for PtpA D126A: a similar K m value and a marked decrease in k cat (36-fold) 21 . Furthermore, the obtained K m values of our mutant were comparable to those reported for other PTPs, particularly the PtpA homologue in Streptomyces coelicolor A3(2) 28 . In contrast, Bach et al found that the D126A mutation caused a total loss of PtpA phosphatase activity, and not a partial loss of function as is expected for this mutant, but still it allowed identifying VPS33B as a putative PtpA substrate 7 . Differences with our results could be due to distinct experimental procedures used to express and/or to purify this mutant. Overall, we obtained a good yield of mycobacterial PtpA D126A mutant that exhibited structural and functional features to act as a useful substrate-trapping mutant for isolating eukaryotic substrates of Mtb PtpA.
Study by SPR of PtpA D126A interactions with macrophage protein components. PtpA D126A was covalently immobilized on a CM5 sensor chip to assess by SPR its ability to bind specifically to macrophage components. For that purpose we prepared a macrophage extract using orthovanadate (Na 3 VO 4 ) and iodoacetic acid (IAA) to preserve P-Y modifications in proteins. Short-time association-dissociation sensorgrams are shown in Fig. 2A. A significant response was detected during the association phase followed by a slow dissociation, suggesting that the extract contains molecules capable of interacting with PtpA D126A. Since these sensorgrams are comparable to those obtained for PtpA wt (inset Fig. 2A), the PtpA D126A mutant seems to be useful to capture potential PtpA substrates present in this macrophage extract. When two successive short pulses of 1.0 M NaCl were injected, the dissociation rate and signal remained similar to that observed with the running buffer demonstrating that the inter-  action was resistant to a high ionic strength ( Supplementary Fig. 3A). This is in agreement with data showing that several hydrophobic residues in the PtpA active site play a key role in the definition of PtpA substrate specificity 24 . In contrast, pulses of 10 mM glycine pH 2.0 caused a blunt drop of the signal to the baseline, indicating the disruption of putative PtpA D126A complexes at low pH ( Supplementary Fig. 3B). The response level was not affected after 4 cycles of regeneration with glycine pH 2.0 ( Supplementary Fig. 3C). Furthermore, Na 3 VO 4 , a competitive inhibitor of tyrosine phosphatases, was utilized to analyze whether the active site was involved in the interactions between PtpA D126A and macrophage components. When the inhibitor was mixed with the macrophage extract before injection, a dose dependent decrease in the total response units (RU) was observed (Fig. 2B). This result suggests that the macrophage extract contained components which were capable of binding to PtpA mutant through the active site. Alternatively, PtpA complexes with macrophage components were allowed to form and then pulses of the inhibitor (25 mM) were applied. In this case Na 3 VO 4 was unable to promote protein components dissociation of PtpA D126A suggesting that the interaction of the PtpA mutant with macrophage components was of higher affinity than with the Na 3 VO 4 . Overall, SPR analysis resulted in a useful tool to define adequate experimental conditions to isolate proteins specifically bound to PtpA D126A during the substrate trapping assay. High salt concentration seems to be a good choice for washing steps to reduce non-specific binding, while low pH would be useful for eluting the proteins bound to PtpA D126A with high affinity.
Substrate trapping assay. On the basis of SPR studies we designed a substrate trapping assay to pull down novel PtpA substrates from the macrophage extract in which P-Y modifications were preserved. The mutant enzyme was covalently immobilized to an NHS-activated sepharose matrix through the same chemistry used in SPR experiments. As a control we verified that the phosphatase activity was not altered after immobilization ( Supplementary Fig. 2C). The binding of PtpA D126A is expected to be uni or bi-punctual since PtpA has only one Lys in addition to the N-terminal group ( Supplementary Fig. 2D). The remaining active groups of the matrix were blocked with ethanolamine, an indispensable step to minimize unspecific protein binding. This constitutes an advantage with respect to the His-tag affinity matrix, used in some cases in substrate trapping experiments. The immobilized PtpA D126A was then incubated with the macrophage extract and subsequently washed with high ionic strength solutions (0.5 M and 1 M NaCl) in order to reduce unspecific protein interactions. The putative PtpA D126A partners were firstly recovered by elution with Laemmli sample buffer. Figure 3 shows the SDS-PAGE analysis of the wash and elution samples obtained in PtpA D126A substrate trapping assays. Following six washes of the matrix, almost no protein was detected by silver staining of the gels (lane w 6 ). After elution, a band pattern (E ST lane) visibly different from that observed for the input macrophage extract (lane MPE) and washes (w2-w6 lanes), was observed. This suggests that potential partners of PtpA D126A were enriched during the substrate trapping assay. In mock substrate trapping experiments, where the macrophage extracts were loaded on a matrix without immobilized PtpA D126A, only few bands were detected in the elution step (lane Ec). All the E ST and Ec lanes were cut and analyzed by MALDI-TOF MS. Using the PtpA D126A immobilized matrix for substrate trapping, a total of eleven proteins were identified from eight gel pieces of the lane E ST (lettered a-h in Fig. 3), while no proteins were identified in the remaining pieces of the gel. The identified proteins are shown in Table 1. In the mock control experiments (Fig 3, lane Ec) we could not identify any protein by MALDI-TOF MS. However, by using a more sensitive MS approach (nano-LC MS) four of the eleven proteins were identified in the mock control experiments as well (HS90, TBB5, TBA1C, ACTB, Supplementary Table 1). Although we cannot exclude the possibility that some of these four proteins have been specifically enriched during substrate trapping, they were excluded for further analysis as they also bind to the PtpA D126A-free matrix. Moreover, these proteins have been previously found to unspecifically bind to different matrixes 29,30 . Taken together these results show that various macrophage components linked to cell energy metabolism were captured using a substrate trapping with PtpA D126A under stringent conditions, including five  mitochondrial (ECHA, ATPA, SQRD, MPCP and CY1) and two cytoplasmic proteins (K6PP and HSPB1) ( Table 1).
As an alternative approach, proteins associated to PtpA D126A during substrate trapping were eluted by lowering the pH to 2.5. As expected, elution in this condition led to the recovery of minor amounts of proteins than when using Laemmli sample buffer, reason why no bands were detected in gels stained with silver nitrate. Thus, these samples were digested with trypsin and the resulting peptides were identified by nano-LC MS analysis. Due to the increased sensitivity of this approach, the total number of proteins identified was greater but included the proteins identified by MALDI-TOF MS after elution with Laemmli sample buffer (Table 1), as well as other subunits of some of these proteins (ECHB, ATPB, ATPO, ATP5H, Supplementary Table 2).
Independently of the experimental approach used for elution and protein identification, from a total of five biological replicates of substrate trapping (including different batches of purified PtpA D126A and macrophage extracts) only four macrophage proteins were systematically identified and thus considered as PtpA putative substrates: ECHA, SQRD, ATPA and K6PP (Table 2). These proteins were identified with the best scores (from 531 to 2422) and 154 peptide spectrum matches corresponding to 69 different peptide sequences (Supplementary Table 3). On the other hand, these candidates were isolated from macrophage extracts where P-Y modifications were preserved by reversible and irreversible inactivation of PTPs with Na 3 VO 4 and IAA, respectively. It is known that in the absence of Na 3 VO 4 the level of P-Y in protein cell extracts is undetectable 31 . However, the use of IAA seems to be of significant help, since we found that the level of P-Y in extracts from Na 3 VO 4 and IAA treated-macrophages was approximately 29% higher than that detected when the P-Y was preserved only with Na 3 VO 4 (Supplementay Fig. 4A). Accordingly, when substrate trapping assay was performed with a macrophage extract prepared without IAA, only two of the four candidates were isolated, and identified with a much lower score (ECHA, score 62 and SQRD score 266). This indicates the relevance of the irreversible inhibition of the endogenous PTPs for isolation of P-Y substrates.
As described above PtpA D126A was barely active (see Supplementary Fig. 2), so it is possible that after substrate trapping the isolated proteins were not totally dephosphorylated. In agreement, the analysis by Western blot using an anti P-Y antibody demonstrated the presence of P-Y residues in the proteins recovered by substrate trapping assay (Supplementary Fig. 4B). However, no P-Y containing peptides were detected by MS analysis. A number of factors usually complicate phosphopeptide identification in complex mixtures using nHPLC-MS/MS approach. During MS/MS analysis the release of phosphate group is the main fragmentation at the expense of peptide bond cleavage, resulting in reduced-quality MS/ MS spectra and lower-confidence in spectral matching for phosphopeptides 32,33 . As shown in Table 3, all the identified PtpA putative substrates contain Tyr residues in the overall sequence and some of them are noted as phosphorylated in the online resource PhosphoSitePlus 34 .
PtpA interacts with and dephosphorylates the TFP (ECHA/ ECHB) in vitro. Using immobilized PtpA D126A we isolated the two subunits of the TFP protein (ECHA/ECHB) as putative substrates of PtpA during substrate trapping experiments. To confirm this protein-protein interaction by a contra-experiment, the purified TFP was covalently immobilized on a SPR sensor chip to assess by SPR its ability to bind specifically to PtpA. The association-dissociation sensorgrams shown in Fig. 4A confirmed that PtpA D126A was capable of interacting with TFP and also with PtpA wt (Supplementary Fig. 5). Furthermore, when the competitive inhibitor Na 3 VO 4 was mixed with the phosphatase before injection, a dose dependent decrease in the total response units was observed (Fig. 4A) suggesting that TFP-PtpA interaction involved the phosphatase active site.
To examine the possibility that phosphorylated TFP (ECHA/ ECHAB) acts as a substrate of PtpA, we tested the ability of PtpA wt to dephosphorylate it. Firstly, TFP was purified by immunoprecipitation using a TFP (ECHA/ECHB) antibody (Fig. 4B) and the presence of P-Y was demonstrated by Western blot using an anti-P-Y antibody (Fig. 4C). Equal amounts of the immunoprecipitated TFP were electro-transferred to a membrane and then incubated with recombinant PtpA wt or buffer as a control. The P-Y signal was then evaluated with anti-P-Y antibody. In comparison with controls, a gradual reduction in the phosphorylation levels of both alpha and beta subunit of TFP was observed after incubation with PtpA wt (Fig. 4C). Altogether, these data indicate that, at least in vitro, TFP is a bona fide substrate of PtpA.

Discussion
This work describes an original methodological approach to improve the substrate trapping assay as a tool to capture potential substrates of phosphatases. This approach is mainly based on the analysis by SPR of the interaction between the enzyme and the extract used as a source of potential substrates, which allows selecting stringent conditions for washing and elution steps, thus reducing unspecific protein interactions. In particular, this methodology was applied to the  identification of novel substrates of Mtb PtpA in human macrophages. For that purpose, we employed the Mtb PtpA D126A mutant and extracts of human THP-1 macrophages were P-Y modifications were preserved. The Mtb PtpA D126A mutant was expressed, purified and characterized, verifying that it mostly retained the biochemical structural properties of PtpA wt while exhibited enzymatic properties suitable to act as a substrate trapping enzyme. The preservation of P-Y modifications in macrophage proteins was assured by using reversible and irreversible phosphatase inhibitors during macrophage extract preparation. As a result of this improved methodology we identified, as novel potential physiological substrates of Mtb PtpA, four macrophage proteins, all linked to cell energy metabolism: TFP (ECHA/ECHB), SQRD, ATPA and K6PP. VPS33B and the GSK-3, previously identified as substrates of PtpA 7 , were not captured in our substrate trapping assay. This contrasting result should be carefully analyzed because of substantial technical differences between these studies and our work. The VPS33B was isolated by substrate trapping using a PtpA carrying the same D/A mutation but exhibiting no phosphatase activity, a non-desired feature for a D/A substrate trapping mutant. In contrast, in our hands, the purified PtpA D126A showed a similar K m and a lower (40-fold) k cat than the PtpA wt, which reproduces the functional features reported in a kinetic study of this PtpA mutant 21 . Furthermore, the experimental design used in this work made more rigorous the conditions used during the substrate trapping assay, so that we may have lost components exhibiting lower affinity for PtpA. With respect to GSK-3, it was found to act as a PtpA substrate using a completely different approach where authors performed a Kinome analysis of pre-selected eukaryotic kinases 15 . Thus, the fact that VPS33B and GSK-3 were not captured in our substrate trapping assay is likely a consequence of different experimental approaches and does not invalid the potential of the molecules identified in this work as PtpA substrates. Moreover, one of them, the TFP, was successfully validated by in vitro experiments.  TFP, ATPA and SQRD, identified as potential Mtb PtpA substrates, are proteins synthesized on cytosolic ribosomes and then translocated to the mitochondria 35,36 . Interestingly, specific and marked changes in mitochondrial ultrastructure and function have been recently described in THP-1 macrophages infected with the virulent Mtb H37Rv strain 37 . Although the molecular events associated with these alterations have not been elucidated yet, significant changes in the mitochondrial proteome were found. Among others, ECHA, ATPA and SQRD were specifically and strongly modulated (more than 10-fold decrease) by the virulent Mtb H37Rv, and, moreover, were no longer detected in the mitochondria 37 . Since PtpA has been localized in the cytosol of Mtb infected macrophages 10 , it cannot be ruled out that it may dephosphorylate these proteins altering their activity and/or translocation to the mitochondria. In any case, this hypothesis requires previous validation of ECHA, ATPA and SQRD as PtpA substrates. As a first step in this direction, we demonstrated that TFP (ECHA/ECHB) is a bona fide substrate of PtpA in vitro (Fig.  4). TFP plays a key role in b-oxidation of long chain fatty acids, a pathway that provides electrons to the mitochondrial respiratory chain for ATP synthesis 38 . Thus, an eventual TFP mitochondrial deficiency in infected macrophages 37 suggests that Mtb drives macrophage catabolism favoring glycolysis over b-oxidation of long chain fatty acids. This metabolic change may play a role in bacterial survival since the host cellular lipids constitute the primary nutrient source for intracellular Mtb 39 . The effect of PtpA on TFP activity and cell distribution needs to be further studied.
The observation that macrophages infected by the virulent Mtb H37Rv, but not by the avirulent Mtb H37Ra strain, exhibited a pro-nounced increase in the ATP to ADP ratio 37 , renders the ATPA subunit, a key component of the ATP synthase catalytic core domain F1, an interesting target for further studies 40,41 . It is worth to mention that ATP synthase subunits have been reported as targets of several viral proteins acting as pro-viral factors regulating virus replication, transmission and propagation in the host 42,43 . The SQRD also represents an interesting candidate as PtpA substrate since this enzyme participates in cell metabolic and microbicidal pathways in cells [44][45][46][47] . This enzyme catalyzes the oxidation of sulfide species (released from the metabolism of sulfur-containing amino acids) to elemental sulfur. The electrons from sulfide oxidation are transferred to the mitochondrial respiratory chain for energy production 45 . Moreover, sulfide oxidation by SQRD contributes to keep sulfide concentration below toxic levels, otherwise sulfide inhibits the cytochrome oxidase (mitochondrial complex IV) interrupting the respiratory chain 46 . On the other hand, in inflammatory macrophages sulfide accumulation likely contributes to down-regulate anti-microbial and pro-apoptotic effects of nitric oxide, as a result of the reaction of sulfide with peroxinitrite 47 . Thus, modulation of SQRD activity and/or cell distribution by Mtb may influence both energy metabolism and nitric oxide-dependent microbicidal effects of macrophages.
Finally, since in vitro and in vivo studies indicate that Mtb infection drives strong metabolic changes in macrophages at the level of glucose consumption 48 it is worth to study whether PtpA is capable of dephosphorylating the cytoplasmic 6-phosphofructokinase (K6P) which constitutes the most important control step in the mammalian glycolytic pathway 49,50 . Indeed, K6P regulates the rate of glycolysis in response to cell energy requirements and is likely linked to the mitochondrial ATP synthesis. Phosphorylation of K6P, which seems to occur mainly in its C-terminal regulatory domain 50 , is involved in the equilibrium between oligomeric forms of the enzyme, in the affinity to substrates and allosteric ligands, in the modulation of the interactions with other proteins, and in its intracellular distribution [50][51][52][53] .
All together, in this work we identified novel components linked to macrophage bioenergetics, particularly TFP, which may be subjected to Mtb modulation through PtpA dephosphorylation. Whether Mtb PtpA-mediated dephosphorilation of these potential targets affects their activity, cell distribution and/or the metabolic pathways of energy production in macrophages requires further analysis.

Methods
Production of recombinant PtpA wt and PtpA D126A and kinetic characterization. The coding sequence of PtpA was amplified by PCR from cosmid MTCY427, using appropriate primers with NdeI and HindIII restriction sites. Digested PCR products were ligated with phage T4 DNA ligase (Biolabs) in a pET28 expression vector. The construct expressing the single PtpA mutant D126A was obtained by site-directed mutagenesis using the Quikchange Site-Directed Mutagenesis kit (Stratagene) and the construct verified by DNA sequencing. Transformed E. coli BL21 (DE) cells were grown at 15uC in LB medium with 50 mg ml 21 kanamycin and protein synthesis induced with 0.5 mM IPTG. The recombinant proteins were purified to homogeneity by metal-affinity (Cu-column) and size exclusion chromatography (SEC). The phosphatase activity was measured using the artificial substrate p-nitrophenyl phosphate (pNPP) at 410 nm 9 and expressed as mM of pNP min 21 (e 5 7.976 3 10 3 M 21 ?cm 21 ). The kinetic constants K m and V max were calculated using the best fit curve to the Michaelis-Menten equation 54 . For more details see legends of Supplementary Fig 1 and 2.
Circular Dichroism. Experiments were performed with 25 mM of PtpA D126A or PtpA wt in 20 mM NaH 2 PO 4 (pH 7.4), 20 mM NaCl. Circular dichroism (CD) spectra were recorded using a JASCO J-715 spectropolarimeter equipped with a thermoelectric sample temperature controller (Peltier system). Data were recorded from 300 to 200 nm at 25uC using a quartz cell with a path length of 10 mm. A total of 5 accumulations were taken for each spectrum with a speed of 50 nm/min, with a resolution of 0.2 nm, responses of 1 s and bandwidth of 1.0 nm. Following background subtraction, all the CD data were converted from CD signal (mdegree) into mean residue ellipticity (deg.cm 2 .dmol 21 ). The secondary structure quantifications were performed with K2d software 27 . Thermal stability of PtpA wt and PtpA D126A was studied by CD spectroscopy. The ellipticity at 222 nm was measured from 25uC to 90uC with an increase step interval of 5uC and a 5 min equilibration at each temperature. Also, renaturation experiments (from 90uC to 25uC) were carried out to ensure reversibility. The change in ellipticity at 222 nm was plotted against the temperature to evaluate thermal stability.
Surface plasmon resonance assays. Experiments were done on a Biacore 3000 GE Healthcare instrument. PtpA D126A was immobilized using standard aminecoupling procedures (Amine Coupling Kit, GE Healthcare) on CM5 sensorchips (GE Healthcare). Ligands immobilization was carried in running buffer 10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% P20 and 1 mM DTT). After surface activation with 0.05 M NHS (N-hydroxysuccinimide) and 0.2 M EDC (1-ethyl-3-(3-dimethylaminopropyl) carbodiimide) PtpA D126A and PtpA wt were diluted in 10 mM sodium acetate pH 4.5 and injected onto two different sensor chip flow cells at a flow rate of 5 ml/min to reach a final density of 1600 and 400 RU, respectively. Remaining sites were blocked with 1.0 M ethanolamine pH 8.5. Another flow cell was activated and blocked in the same way with no ligand injections and used as reference surface. Sensorchip was washed with 1.0 M NaCl and stabilized with running buffer before sample injections. Injections of macrophage extracts (1.2 mg in protein) were performed in running buffer at 25uC and a flow rate of 10 ml/min unless otherwise stated. Interactions were evaluated by the analysis of the association/ dissociation sensorgrams, and their stability was assessed by the injection of two 30 second pulses of 1 M NaCl and by the injection of two 40 second pulses of 10 mM glycine pH 2.0. SPR experiments carried out to study the interaction between TFP (ECHA/ECHB) and the phosphatases (PtpAwt and PtpA D126A) were done in the same way by immobilizing TFP on a CM5 sensorchip to a final density of 360 RU (at flow rate 50 ml/min) and remaining sites blocked using the same procedure. Running buffer used in these experiments was 20 mM Tris-HCl pH 8.0, 50 mM NaCl, 5 mM EDTA, 0.005% P20 and 1 mM DTT and phosphatases injected at flow rate 20 ml/min unless otherwise stated. The effect of incubating with increasing concentrations of Na 3 VO 4 (2.5-25 mM) was also evaluated. All data processing was carried out using the BIAevaluation 4.1 software (GE Healthcare).
Substrate trapping using immobilized PtpA D126A. Purified PtpA D126A was covalently coupled to NHS-activated sepharose (GE Healthcare) following instructions provided by the manufacturer. Briefly, the matrix (100 ml) was washed with cold 1 mM HCl, and immediately 300 mg of PtpA D126A diluted in coupling buffer (0.2 M NaHCO 3 , 0.5 M NaCl pH 8.3) were added and incubated 16 hs at 4uC. Unreacted groups of the matrix were blocked overnight at 4uC with 0.5 M ethanolamine pH 8.3, 0.5 M NaCl. Then, the matrix was washed with 0.1 M Tris-HCl pH 8.0 and 0.1 M acetate buffer pH 4.5, 0.5 M NaCl. In parallel, as a control for unspecific binding in pull-down assays the same amount of matrix was incubated in coupling buffer without PtpA D126A, and then blocked with ethanolamine following the same protocol, and this control was defined as the mock substrate trapping (see Supplementary Fig. 2). For each substrate trapping assay, 100 ml of the matrix with or without immobilized PtpA D126A was incubated with 5 mg of macrophage extract diluted to 0.17 mg/ml in SPR running buffer containing 1 mM benzamidine, 1 mM PMSF, 1 mg/ml SBTI during 1 h at 4uC with gentle end-over-end agitation. The matrix was collected by low-speed centrifugation (1,0003 g) and then washed twice and sequentially with 100 ml of running buffer and running buffer containing 0.5 M and 1.0 M NaCl at 4uC. The retained proteins were eluted using one of the two approaches described below. In the first case, 50 ml of Laemmli sample buffer were added to the matrix and then boiled for 5 min. All washes and elutions were analyzed by SDS-PAGE followed by silver staining. This approach was performed on two biological replicates. The lanes corresponding to the eluted proteins were cut in 2 mm bands and each gel piece was analyzed by MALDI-TOF MS. In the second approach, to reproduce the elution conditions of the SPR assays, proteins were eluted with successive additions of 50 ml of 10 mM glycine pH 2.5 followed by an immediate neutralization with 2 ml of 1 M Tris-HCl pH 7.5, and stored at 220uC until analysis by Nano-LC-MS/MS. This approach was performed on three biological replicates. Each biological replicate was performed using a different batch of the macrophage extract and purified PtpA D126A.
Mass spectrometry. For MALDI-TOF MS analysis proteins were in-gel digested with trypsin (sequence grade, Promega) as previously described 56 . Peptides were extracted from gels using aqueous 60% acetonitrile (ACN) containing 0.1% TFA and concentrated by vacuum drying. Prior to MS analysis, samples were desalted using C18 reverse phase micro-columns (Omix H Tips, Varian) and eluted directly onto the MALDI sample plate with matrix solution a-cyano-4-hydroxycinnamic acid in 60% ACN containing 0.1% TFA. Mass spectra of peptides mixtures were acquired in a 4800 MALDI TOF/TOF instrument (ABi Sciex) in positive reflector mode and were externally calibrated using a mixture of peptide standards (Applied Biosystems). Collision-induced dissociation MS/MS spectra of selected peptides ions were acquired. Proteins were identified with measured m/z values in MS and MS/MS acquisition modes and using the MASCOT search engine (Matrix Science, http:// www.matrixscience.com) in the Sequence Query search mode. The following search parameters were used for searching the NCBInr database (NCBInr 20130721): taxonomy Homo sapiens; protein mass was unrestricted; monoisotopic mass tolerance, 0.05 Da; fragment mass tolerance, 0.3 Da; partial methionine oxidation, cysteine carbamidomethylation and tyrosine phosphorylation as variable modifications; and one missed tryptic cleavage allowed. Significant protein scores (p , 0.05) and at least one peptide ion significant score (p , 0.05) per protein were used as criteria for positive identification.
For nano-LC-MS-analysis, proteins in bands were in-gel digested as described above. In the case of liquid samples (50 ml), proteins were digested with sequencinggrade trypsin (0.25 mg, 12 h at 37uC), desalted, dried by vacuum and resuspended in 20 ml of 0.1% formic acid (v/v) in water. Samples were injected into a nano-HPLC system (Proxeon easynLC, Thermo Scientific) fitted with a reverse-phase column (easy C18 column, 3 mm; 75 mm ID 3 10 cm; Proxeon, Thermo Scientific) and peptides were separated with a linear gradient of acetonitrile 0.1% formic acid (0-60% in 60 min) at a flow rate of 400 nL/min. Online MS detection/analysis was carried out in the LTQ Velos nano-ESI-linear ion trap instrument (Thermo Scientific) in a datadependent mode (full scan followed by MS/MS of the top 5 peaks in each segment, using an exclusion dynamic list). Proteins were identified by searching the SwissProt database (November 2012), taxonomy Homo sapiens, using the following parameters in the MS/MS ion search mode of Mascot search engine: peptide tolerance 1.5 Da, MS/MS tolerance 0.8 Da, and cysteine carbamidomethylation, methionine oxidation and tyrosine phosphorylation as the allowed variable modifications. The significance limit for protein identification was set at p , 0.01 and an ion cut off . 40. Only proteins identified with two or more peptides were considered positively identified. The list of potential PtpA partners was elaborated by manually removal of all the proteins identified in mock control experiments. Only proteins identified in all substrate trapping (five biological replicates), independently of the approach of protein elution and MS analysis, were considered as potential substrate.