We have previously reported that recombinant PtpA (rPtpA) interacts and dephosphorylates hTFP_α immunopurified from macrophages and that its dephosphorylation is dependent on the PtpA dose (Margenat et al., 2015). To identify the phosphotyrosine potentially dephosphorylated by PtpA, we performed in silico docking studies, using the crystal structure of hTFP_α (extracted from PDB 6DV2) and of Mtb PtpA (PDB 1U2P) (Madhurantakam et al., 2005; Xia et al., 2019). The reported structure of the hTFP (Xia et al., 2019) showed that, as its bacterial homolog (P. fragi, PDB 1WDK) (Ishikawa et al., 2004), the biological unit was a α ₂ β ₂-heterotetramer. Nevertheless, hTFP presents significant primary and secondary structure differences compared to its bacterial counterparts, some of them in the α subunit, described as relevant for its association with the inner mitochondrial membrane (Fould et al., 2010; Xia et al., 2019). To drive the docking process, we selected the PtpA active site as we previously showed by SPR assays that the PtpA-hTFP interaction involves the active site of PtpA (Margenat et al., 2015). For hTFP_α, we selected all tyrosine residues located at the surface (Tyr43, Tyr239, Tyr271, Tyr435, Tyr637, Tyr639, Tyr724, and Tyr762), and we phosphorylated them in silico. The inset table ( Figure 1A ) shows the different numbers of complexes (PtpA/hTFP_α) and clusters obtained for each p-Tyr evaluated in the docking assays. The first four best representative complexes from each cluster were selected, counting 256 possible solutions. It is worth observing that a p-Tyr, with a lower number of clusters, indicates that the binding mode is well-defined and better than another with a greater number of clusters. Additionally, the clusters with lower electrostatic interaction energies (represented by HADDOCK scores) or with a higher number of members are preferable. Considering this, the interaction through the p-Tyr271 or p-Tyr639 of hTFP_α and the active site of PtpA results in at least one complex with negative electrostatic interaction energy. However, only the cluster of hTFP_α phosphorylated at the Tyr271 residue has a negative average electrostatic interaction energy. In this case, all complexes within this cluster have negative electrostatic interaction energy, and this cluster represents the one with the least energy and the most populated. The situation is different for the results of hTFP_α interaction through p-Tyr639. In this case, we found that the cluster with a higher number of members differs from the one with lower energy. Only one complex in the last has negative electrostatic interaction energy. The best PtpA-hTFP_α complex, involving Tyr271, is shown in Figure 1B . In this protein complex, it is possible to observe that the p-Tyr-271 fits in the PtpA active site, and there are no steric clashes between both proteins. The p-Tyr271 of hTFP_α is directly interacting with the catalytic Asp126 and the Cys11 and Cys16 of PtpA, as it should be for the enzymatic reaction to occur (Madhurantakam et al., 2008).

In addition, we evaluated if there exists a common sequence or structure pattern among the reported PtpA targets. For the mycobacterial substrate of PtpA, the ATP synthase subunit alpha (ATPA), the p-Tyr targeted by PtpA still needs to be elucidated (Chatterjee et al., 2019). For the eukaryotic substrates, GSK3α and VPS33B, only indirect evidence of which Tyr could be dephosphorylated by PtpA has been reported (Bach et al., 2008; Poirier et al., 2014). When considering these residues (Tyr279 of GSK3α, Tyr133 or Tyr382 of VPS33B) and the available alphafold structures models of these substrates, we did not find any sequence pattern around those residues ( Figure S1A ) or a conserved structural motif of interaction between PtpA and the reported substrates. One reason that could explain this result is that the p-Tyr of GSK3α and VPS33B were not identified using the same approach as our study, where an unbiased in silico analysis was applied to determine the p-Tyr candidates dephosphorylated by PtpA. Therefore, we cannot rule out that other Tyr of GSK3α and VPS33B could be the target of PtpA. Another reason is that alphafold models of GSK3α, VPS33B, and ATPA did not consider the phosphorylated state of the residues. Therefore, the orientation of the side chain of these residues might not be the correct one in the alphafold-modeled structure. Overall, we think that an unbiased in silico study, as carried out for PtpA with the hTFP_α as substrate, is needed for the other reported substrates to be sure of which Tyr is potentially dephosphorylated by PtpA and to evaluate correctly if a conserved recognition motif exists.

As it has been proved that mycobacterial PtpA introduced into the cytosol of the infected cells acts principally on eukaryotic proteins, this enzyme is expected to dephosphorylate a specific tyrosine residue of hTFP_α, absent or not phosphorylated in the bacterial homologs. Thus, we evaluate the conservation of the Tyr271 and Tyr639 in bacterial TFP_α and eukaryotic homologs ( Figure 2A , see Figures S1 – S3 for the complete alignments). While Tyr639 is in the HACD catalytic domain and is conserved in mammals and bacteria, Tyr271 seems specific to the mammal’s group, and it is part of the helix-10 of hTFP_α which is absent in bacterial TFP homologs ( Figures 2A, B ). Interestingly, this helix-10 was described as one of the helices relevant for the activity and mitochondrial localization of the hTFP (Xia et al., 2019). Then, we analyzed in detail the conservation of Tyr271 in eukaryotes employing blastp searches of the hTFP against the NCBI RefSeq protein database and a maximum likelihood phylogenetic analysis ( Figure 2B ). In chordates, Tyr271 is mainly conserved in mammals, birds, reptiles, and amphibians ( Figures 2B , S3 ). In fish, there are two major clades; one contains 146 genes with a high degree of Tyr271 conservation, and the other with 57 genes without conservation of Tyr271. A closer look at this non-conserved clade indicates that these genes are, in most cases (53/57), different isoforms of a gene containing a conserved Tyr271. Fishes are generally polyploid (Rajkov et al., 2014), and additional copies of duplicated genes are free to change regarding one of the copies remaining functional. On the contrary, there is no conservation of Tyr271 in non-chordates (like arthropods, mollusks, or cnidaria), suggesting that the preservation of the Tyr271 is a characteristic of more complex organisms. These results allowed us to hypothesize a potential role of the Tyr271 residue in the regulation of hTFP_α subcellular localization/activity, supported by the fact that the subcellular localization of hTFP_α was modulated during TB infection, where this subunit was no longer detected in the mitochondria of macrophages infected with the virulent Mtb H37Rv (Jamwal et al., 2013).

The next step was to show if Tyr271 of hTFP_α could exist in its phosphorylated state. The hTFP_α contains several post-translational modifications (PTMs) identified after large-scale MS studies and noted in the PhosphoSitePlus (Hornbeck et al., 2012). Among them, nine Tyr residues of hTFP_α were reported as phosphorylated, but the Tyr271 is not included ( Figure S7 ). The functional significance of p-Tyr and the kinase involved has not been elucidated yet. In this context, to improve the identification of the p-residues of hTFP_α, we prepared a batch of immunopurified hTFP from a macrophage phosphoprotein-enriched extract, as previously described (Margenat et al., 2015). We verified in this sample the presence of hTFP_α by MS ( Figure S4A ). However, the amounts of immunopurified hTFP were insufficient to detect the p-Tyr by MS and evaluate PtpA-mediated p-Tyr271 dephosphorylation. Phosphorylation generally represents a lower percentage of the total protein and could be lost during MS fragmentation (Mann et al., 2002; Dephoure et al., 2013) or by the action of endogenous phosphatases during the protein extract preparation, despite the precautions taken to inhibit them. This explanation is supported by the fact that Tyr271 is exposed to the solvent, as seen in its crystallographic structure ( Figure 1B ). Thus, to obtain enough material, we decided to produce and purify the recombinant hTFP_α based on previous evidence of its successful production in a soluble form in E. coli (Fould et al., 2010). After IMAC, SEC, and removal of the His-tag, 2.5 mg of soluble hTFP_α/gr of E. coli pellet was obtained. As illustrated on the SEC chromatogram, we observed the main peak with a shoulder aspect ( Figure 3A ). A pool of the eluting fractions corresponding to this peak was analyzed by SDS-PAGE and MS, showing that the principal band is the hTFP_α, and the less intense band corresponds to an E. coli chaperone contaminant ( Figure S4B ). Then, taking into account that it is not known which eukaryotic kinase is involved in its phosphorylation, we decided to evaluate the Tyr phosphorylation by Janus kinase (Jak). This enzyme is an intracellular kinase involved in cytokine-mediated signaling, for which inhibitors of its activity were proposed for host-directed therapy of TB as an adjunct to standard TB chemotherapy (Dorhoi, 2015; Schwartz et al., 2017). Thus, we incubated the recombinant hTFP_α with catalytic amounts of the Jak kinase domain, using an E:S molar ratio of 1:3000 to avoid non-specific phosphorylation. For the recombinant hTFP_α non-treated with Jak, after SDS-PAGE and WB using an anti-p-Tyr antibody, we detected a signal of Tyr phosphorylation of hTFP_α, and by MS, we identified Tyr499 as phosphorylated ( Figure S6 , already reported in PhosphoSitePlus). This result is explained because the recombinant hTFP_α was expressed in an E. coli strain with a tyrosine kinase/phosphatase system (Vincent et al., 1999). Then, for the recombinant hTFP_α pre-treated with Jak and ATP, we observed additional phosphorylation, detected by the electrophoretic shift of the protein in the SDS-PAGE (Villarino et al., 2005) as well as by WB and MS analysis ( Figures S5 , S6 ). Using this strategy, we identified two new sites of phosphorylation of hTFP_α, the p-Tyr271 of helix-10 ( Figure 3B ) and the p-Tyr43 of the predicted signal peptide; and the p-Tyr435 of HACD already reported in the PhosphoSitePlus. This result is relevant because until now, the phosphorylation sites Tyr43 and Tyr271 and the potential kinase involved in hTFP_α phosphorylation were not described. Figure S7 summarizes all the PTMs of hTFP_α reported up to now.

First, we characterized the recombinant protein hTFP_α by SEC. We determined its apparent protein molecular weight noting that hTFP_α eluted as a protein dimer (179 kDa) of two alpha subunits ( Figure 3A ). We suggest that this dimer is stabilized by hydrophobic regions of the alpha subunits, which in the biological unit of hTFP (α ₂ β ₂-heterotetramer) interact with the beta subunits (Xia et al., 2019). We also produced with a high degree of purity and yield, the recombinant hTFP_α-Y271F mutant, which, as the hTFP_α-wt protein, is a dimer in solution ( Figure S8 ). For the PtpA activity assay both hTFP_α and hTFP_α-Y271F were previously phosphorylated by Jak. The recombinant PtpA-wt and the inactive mutant PtpA-C11S were also produced as described previously (Margenat et al., 2015), and characterized by SEC as monomeric proteins ( Figure S9 ). Then, we evaluated PtpA activity using a method previously reported (Najarro et al., 2001), based on the incubation of the phosphatase with the protein substrate previously separated by SDS-PAGE, transferred and immobilized to a membrane in a non-covalent way. After immobilization, the hTFP_α substrate was expected to recover part of its secondary structure, because the denaturing agent was removed. Using this method we verified the p-Tyr phosphatase activity of PtpA ( Figure S10A-D ). However, this strategy was not appropriate to evaluate the substrate specificity of PtpA, because the substrate did not preserve its global structure needed to a correct enzyme-substrate recognition. In this context, we observed that PtpA dephosphorylated p-Tyr residues no matter which immobilized protein was used (TFP or the viral protein OH1-C112S) ( Figure S10E ). On the other hand, carrying out the activity assay in solution presents difficulties, as a detergent is needed to maintain the recombinant hTFP_α protein in solution, since this protein is described as associated to the mitochondrial membrane (Fould et al., 2010; Liang et al., 2018; Xia et al., 2019). High concentrations of detergent inhibit PtpA and interfere with the commonly used Malachite Green reagent in the Pi detection. Thus, it was necessary to define the optimal conditions of the reaction to preserve the hTFP and PtpA stability during the incubation at 37°C (see methodology). In this assay we used catalytic amounts of PtpA (1.2 nM) corresponding to an E:S molar ratio of 1:5000, and observed a significant PtpA dephosphorylation of hTFP_α-wt but not of hTFP_α-Y271F ( Figure 3C ). This result can be explained by the absence of p-Y271 in the mutant, suggesting that this residue is the preferred target of PtpA in vitro. When the assay was performed with a higher concentration of PtpA (12 nM, E:S molar ratio of 1:500) the dephosphorylation of hTFP_α-Y271F started to be detected ( Figure S11 ). In this condition, it is not possible to avoid dephosphorylation of other p-Tyr residues that probably do not represent specific targets of PtpA. Then, to confirm the dephosphorylation site/s of the recombinant hTFP_α we performed nanoLC-MS/MS analyses. Nevertheless, despite the addition of phosphopeptide enrichment steps, we could not confidently detect phosphorylation sites in hTFP_α to perform a quantitative analysis of dephosphorylation by PtpA. More long-term research efforts to improve and optimize the detection of phosphosites in this protein are needed. Finally, to evaluate PtpA specificity, we used the recombinant hTFP_α and the inactive mutant of the viral protein OH1 (OH1-C112S) as a substrate (Segovia et al., 2017), previously phosphorylated by Jak kinase. As shown in Figure 3D we observed a significant decrease of the p-Tyr signal when hTFP_α was used as a substrate but not using OH1-C112S (120 nM of PtpA, E:S molar ratio of 1:50), demonstrating the specificity of PtpA for hTFP_α.

We have already observed by SPR studies that PtpA interacts through the active site with hTFP immunopurified from macrophage (Margenat et al., 2015). In the present study, we performed additional SPR studies to determine the affinity constant of the interaction of the complex, whose value was 0.31 μM ( Figure S13 ), a value similar to that reported for other phosphatases and their substrates (Czikora et al., 2011). Previously to this assay, we verified that 21% of the immunopurified hTFP is phosphorylated on Tyr, using an anti-p-Tyr-Ab ( Figure S10A ). In addition, we performed molecular dynamics simulations (MD) to determine features of the interaction between PtpA and hTFP_α, using the best complex obtained by docking described above. Three systems were considered for MD simulations in solution: PtpA, hTFP_α, and the PtpA-hTFP_α complex. For the last, three sets of independent simulations (with different initial random Maxwell-Boltzmann velocities) were performed. According to the Root Mean Square Distance (RMSD) of the backbone atoms along the nanoseconds of simulation, both proteins in the complex maintained their structure stable, as they are alone in solution ( Figure 4A ). This result suggests that the complex is stable and remains formed over the time analyzed. Concerning the Root Mean Square Fluctuation (RMSF) values, we observed that hTFP_α fluctuation in the complex is similar (on average) to the protein alone. In contrast, PtpA fluctuations in the complex are reduced compared to when the protein alone is evaluated ( Figure 4B ). The last suggests a change in PtpA behavior when the complex is formed. Interestingly, the regions corresponding to the active site of PtpA have a more considerable relative decrease in their fluctuation when the complex is formed. For instance, the region that corresponds to the D-loop (Asp126) of PtpA is stabilized upon the PtpA-hTFP _α complex formation, probably reflecting the conformational change that allows the catalytic Asp to adopt the correct position during catalysis (Madhurantakam et al., 2008; Hobiger and Friedrich, 2015).

As described in the introduction, some reports connect PtpA with the degradation pathways induced by ubiquitin. It has been reported that PtpA binds non-covalently host ubiquitin via a previously unknown ubiquitin-interacting motif-like (UIML) and that this interaction activates PtpA to dephosphorylate the kinase Jnk and the p38 MAPK, which leads to the suppression of innate immunity (Wang et al., 2015). In this context, we decided to characterize the ubiquitin:PtpA interaction by molecular docking and MD and perform kinetic studies to evaluate the proposed modulation of PtpA activity. The interacting configuration between ubiquitin and PtpA was determined using the HADDOCK web server (Van Zundert et al., 2016). For that, the residues from the UIML region of PtpA and the residues of ubiquitin proposed as relevant for the interaction (His 68, Ile 44, and Leu 8) were selected (Wang et al., 2015). The best result was obtained from the cluster with the lowest HADDOCK score and also the most populated one. The interacting region has a hydrophobic core (HC) composed of residues from the UIML region of PtpA (Ala 140 and Val 141) and ubiquitin residues (Leu 8 and Val 70) (indicated in Figure 5A ), flanked by hydrogen bond interacting residues, Glu 143 (PtpA) with Arg 42 (ubiquitin), Glu 137 (PtpA) with His 68 (ubiquitin), and Arg 111 (PtpA) interacting with the backbone oxygen of Leu 8 (ubiquitin). To test the stability of the complex, we performed three parallel MD simulations using as a starting point the interacting complex described above. The MD results show that the HC is stable and remains conserved in all three simulations while the surrounding hydrogen bonds form, break down, and change over time. In particular, the final configuration of the complex is different in the simulations, as shown in Figure 5B . In the first MD, the ubiquitin rotates around the HC and forms different H-bonds in comparison with the starting point during the MD simulation. In the second case, the interactions between both proteins are stable and the same as at the beginning of the MD. Finally, in the third MD the ubiquitin moves and rotates over the HC, changing all H-bonds interaction in comparison with the initial configuration and the relative position of the Val 70 residue; nevertheless, the HC is still formed. This result suggests that ubiquitin can interact with PtpA during the simulation, albeit with nonspecific interactions and with the HC anchoring the interaction. Furthermore, the RMSD along the trajectory for both proteins of the complex shows values of less than 3 Å ( Figure S14A ), similar to the values for the proteins alone in solution, implying that both keep their respective internal conformation along the trajectory, without additional stabilization.

To check whether ubiquitin could affect PtpA activity, we evaluated putative changes in the PtpA active site during the interaction. To this end, we performed a simple geometric approach, in which the surface of the triangle formed by three relevant heavy atoms of the PtpA active site was followed along the trajectory (S atom of Cys 11 and Cys 16 together with a C atom of the Asp 126 lateral chain) when PtpA is alone or interacting with ubiquitin ( Figure 5B ). This figure shows two possible and stable values for the calculated triangular surface, meaning two possible conformations of the PtpA active site: one with a higher area reflecting an open conformation and the other with a lower area indicating a closed conformation. In the case of the MD simulation of PtpA alone, and only one trajectory of the complex, we found an open conformation, while in two other MDs of the complex, a close conformation was found ( Figure S14B ). However, no correlation between the final position of ubiquitin in the complex and the conformation of the PtpA active site was found, explaining the reported PtpA activating effect of the ubiquitin. In addition, we experimentally tested the impact of ubiquitin on PtpA activity using good-quality recombinant ubiquitin and active PtpA proteins ( Figure S15A ). Contrary to what was previously reported (Wang et al., 2015), we did not detect activation of PtpA acting on the artificial substrate pNPP. We observed the same result, even varying the PtpA:ubiquitin molar ratio from 1:0 to 1:10 ( Figure S15B ). Considering that the authors used the PtpA linked to the GST-tag and we used PtpA linked to a His-tag, we evaluated whether the discrepancy could be due to the His-tag used. Even after removing the tag, no activating effect of ubiquitin on PtpA phosphatase activity was observed ( Figure S15C ). Thus, our discrepancy with the data reported by Wang et al., 2015 could be due to the distinct fusion protein used. Overall, our in silico and experimental results suggest that the interaction of ubiquitin with PtpA cannot explain the reported PtpA activation. However, we cannot discard that additional factors may be required to achieve this effect. In parallel, we also analyzed whether the interaction of PtpA with ubiquitin can occur in the context of the PtpA-hTFP _α complex. This analysis could be interesting because hTFP _α has several ubiquitylation sites noted in the database Phosphosite (Hornbeck et al., 2015), and we demonstrated that it co-purifies with the E3 ubiquitin ligase (TRIM21, Figure S4A ), which could connect hTFP_α with the Ub-mediated proteasomal degradation. Figure 5C shows that PtpA could form the protein complex with ubiquitin without steric clashes. However, our study did not detect activation of PtpA by ubiquitin acting on hTFP_α, after three independent assays ( Figure 5D ). As shown in the graph, we observed a significant decrease in P-Tyr/hTFP_α signal level after incubation with PtpA in the absence or presence of Ub. However, in the case of the assay with PtpA+Ub, this decrease was lower than without Ub. This suggests an inhibitory effect of ubiquitin on the phosphatase activity of PtpA acting on hTFP_α as a substrate, consistently with the detection of an alternation of the active site between an open (active) and closed (inactive) conformation during PtpA/ubiquitin MD simulation ( Figure S14B ). Further studies using the recombinant TFP_α/β heterotetramer as a substrate will be interesting to determine if the inhibitory effect of the ubiquitin is maintained.

The following protocol was used to select which tyrosine may be phosphorylated in the hTFP_α protein, in addition to those reported in the PhosphoSitePlus^® (Hornbeck et al., 2015), and acts as putative interacting residue with mPtpA when hTFP_α is its substrate. The protocol was performed for each tyrosine in hTFP_α located at the surface of the protein, and it consists of three steps involving (i) Selection of the putative phosphotyrosine on the hTFP_α, (ii) docking simulations, and (iii) electrostatics calculations. (i) Selection of putative phosphotyrosine on hTFP_α. The selected pdb structure for hTFP_α was the only X-ray structure available in the PDB databank (Berman et al., 2000), pdb code 6DV2, with a resolution of 3.6 Å (Xia et al., 2019). The putative tyrosines from this structure that could be phosphorylated were selected based on their exposure to the protein’s surface. The selected tyrosines, with at least 40% of their surface exposed to solvent in the pdb structure, were Y43, Y239, Y271, Y435, Y637, Y639, Y724, and Y762. Each of these residues was phosphorylated in silico by hand one by one, adding the necessary atoms to the tyrosine residue. Energy minimization of the complete protein structure was then performed in implicit solvent using the Amber package of programs (Pearlman et al., 1995) with the amber ff14sb force field (Maier et al., 2015). The force field parameters for the phosphotyrosine were taken from the publication by Homeyer (Homeyer et al., 2006). The implicit solvent was taken into account using the modified Generalized Born model developed by A. Onufriev et al. (Onufriev et al., 2004, 2000) as implemented in the amber program (Case et al., 2005). Finally, 8 different hTFP_α structures were obtained, each with one phosphotyrosine. (ii) Docking simulations: The HADDOCK program (Dominguez et al., 2003; Van Zundert et al., 2016) was used for the molecular docking simulations between PtpA and hTFP_α with one phosphotyrosine. The conformation of the PtpA protein was taken from the X-ray structure with a resolution of 1.9 Å with the pdb code 1U2P (Madhurantakam et al., 2005). The active residues at the protein interface for docking simulation were defined as the catalytic residues Cys 11, Cys 16, and Asp 126. For hTFP_α, each phosphorylated tyrosine was treated as an active residue. The passive residues for both proteins were defined as all surface residues within 6.5 Å from the active residues on each protein. Different numbers of clusters were obtained for each docking simulation. Results, around 200 different complexes for each docking simulation were clustered according to the HADDOCK protocol (Dominguez et al., 2003; Van Zundert et al., 2016). (iii) Electrostatic calculations: Since the HADDOCK score cannot be used to compare the results of the hTFP_α protein with different phosphorylated tyrosines, but only to compare different solutions with the same phosphorylated tyrosines for a given complex (Kastritis and Bonvin, 2010), the electrostatic contribution to the free energy of binding of the interaction between PtpA and hTFP_α was used as a criterion for selecting the best representative complex. Therefore, the complex with the lowest electrostatic energy of binding was selected as the best representative. The electrostatic energy was calculated by solving the nonlinear Poisson-Boltzmann equation for the complex and each protein using APBS 3.2.0 software (Jurrus et al., 2018), using a dielectric constant for points within and outside the protein of 2.0 and 78.5 D respectively. The electrostatic energy of binding (E) was then calculated as the difference between the electrostatic energy of the complex, composed of the hTFP_α and PtpA proteins, and the electrostatic energy calculated for both hTFP_α and PtpA proteins individually, E=EhTFP_α+PtpA-(EPtpA+EhTFP_α). We consider a negative value of E as an indicator that the complex is formed and may be stable.

All MD simulations were performed using the Amber package version 2018 (Case et al., 2005). Parameters from the Amber ff14SB force field were used for all protein residues (Maier et al., 2015); while the parameters for the phosphotyrosine were taken from the literature (Homeyer et al., 2006). Periodic boundary conditions were taken into account in all simulations. A weak-coupling algorithm (Berendsen et al., 1984) was used to couple the simulation boxes with an isotropic pressure of 1 atm and a reference temperature of 300 K. Relaxation times were chosen to be 5 ps and 2 ps for pressure and temperature coupling, respectively. All bonds involving hydrogen were constrained using the SHAKE algorithm (Hess et al., 1997). The time step for all simulations was set to 2 fs. All systems were immersed into a truncated octahedron of TIP3P water molecules (Price and Brooks, 2004) with a minimum distance between the edges of the simulation box and the protein surface atom of 11 Å. The bonds of the water molecules were constrained using the SETTLE algorithm (Miyamoto and Kollman, 1992). The complete systems were neutralized with counter-ions. A direct cutoff for nonbonded interactions of 10 Å and the particle mesh Ewald for long-range electrostatics were applied (Essmann et al., 1995). A standard protocol was used for all MD simulations, and it begins with two initial energy minimizations: the first one only for the solvent molecules, constraining the position of all non-solvent atoms, and a second minimization for the entire system. In order to equilibrate the systems at 300 K, each of them was slowly heated from 100 K to 300 K over a period of 1 ns under NVT conditions, followed by MD simulations at 300 K up to 100 ns. Subsequently, a production run of 400 ns or 500 ns was performed for each complex under NPT conditions. For each simulated system, different numbers of parallel runs were performed with different random initial velocities in order to have more statistics.

The human trifunctional enzyme (NCBI accession: NP_000173.2) was used as a query for a blastp search against the NCBI RefSeq database (e-value: 1e-5, number of hits: 5000). Hits were manually filtered to those containing “trifunctional” in their descriptions. When more than one isoform was available, the longest one was maintained. Hits were further filtered to those with a query coverage greater than 50%. Sequences were submitted to NGPhylogeny.fr (Lemoine et al., 2019) for alignment with MAFFT (Katoh and Standley, 2013), alignment curation with BMGE (Criscuolo and Gribaldo, 2010), evolutionary model selection, and phylogenetic analysis with SMS (Lefort et al., 2017) and PhyML (Guindon et al., 2010), respectively. Trees were visualized and annotated using ITOL (Letunic and Bork, 2021).

The plasmid pET28 containing the sequence of PtpA was obtained as described previously (Margenat et al., 2015). The construct expressing the single PtpA mutant C11S was obtained by site-directed mutagenesis (Quikchange Site-Directed Mutagenesis kit, Stratagene) using the following primers forward and reverse, respectively: 5´ -TTC GTT TCT ACG GGC AAC ATC TG-3, -5´- CA GAT GTT GCC CGT AGA AAC GAA - 3´. The sequence was verified by DNA sequencing (Macrogen). Expression was done as described previously (Margenat et al., 2015). Briefly, Transformed E. coli BL21 (DE3) cells were grown at 15°C in LB medium with 50 μg ml^-1 kanamycin, and protein synthesis induced with 0.5 mM isopropyl β-D-thiogalactoside (IPTG). The recombinant proteins were purified to homogeneity by metal-affinity (Cu-column) and size exclusion chromatography (SEC).

The hTFP _α/β was obtained from macrophage extracts by immunoprecipitation with an anti-TFP monoclonal antibody. Briefly, the anti-TFP MAb (8μg, ab110302, MitoSciences) was first covalently cross-linked to anti-mouse IgG Ab on the beads (100 μl, 11201D, Life technologies) using BS3 (Sigma). Then, beads were washed and incubated with 500 μg of macrophage extract, prepared as previously described (Margenat et al., 2015). Beads were washed, and bound proteins were eluted with 50 mM citrate pH 2.6 (2x100 μl), and neutralized immediately. Samples were resolved by SDS-PAGE, stained with colloidal Coomassie, or transferred to PVDF membranes for 1 h at 100 V. Membranes were blocked for 16 hs at 4°C, washed twice with TBS-T, and subsequently incubated with anti-hTFP_α MAb (ab200652) at 1/1000 dilution in TBS-T, 1.5h at RT. After washing, membranes were incubated (1 h at RT) with an anti-rabbit antibody conjugated with horseradish peroxidase (HRP) (1/50000, Sigma-Aldrich 0545). After four washes with TBS-T and one wash with TBS, the reaction was developed with Pierce ECL western blotting substrate (Thermo Scientific). Immunoreactive bands were visualized using the GBOX ChemiSystem tool (SynGene). The presence of the TFP was confirmed by mass spectrometry (MS).

The hTFP_α-wt (ECHA sequence) was sub-cloned by RF-cloning (Unger et al., 2010), in a modified pET32a vector (named pT7), carrying an ampicillin-resistance cassette, an N-terminus His-tag sequence, and the tobacco etch virus (TEV) protease recognition site between this tag and the target gene insertion site (Correa et al., 2014). As a template the ECHA sequence already cloned in the pMyr-plasmid was used. The following forward and reverse primers were used, respectively: 5’-GGATCGGAAAAC CTGTATTTTCAGGGATCCACCAGAACCCATATTAACTATGGAG-3’ and 5’-GAACTGCGGGTGGCTCCAGCTGCCGGATCCTCACTGGTAGAACTTCTTGTTAGG-3’. PCR were done using Phusion Polymerase (Thermo), and the PCR conditions were: a denaturing step at 98°C for 30 s and 35 amplification cycles of 98°C for 10 s, 65°C for 30 sec and 72°C for 1 min 20 sec, with a final extension step at 72°C for 10 min. The obtained products (megaprimer) were analyzed by agarose gel electrophoresis, purified with GeneJET Extraction Kit (Thermo) and quantified in a nanodrop spectrophotometer. The generated megaprimers contained 30 bp in both ends that overlaps with the insertion site in the destination vector T7. The integration into the vector was done by performing a second PCR reaction using 200 ng megaprimer, 30 ng pT7 vector, and the PCR conditions as follows: a denaturing step at 98°C for 30 s, 30 amplification cycles of 98°C for 10 s, 60°C for 1 min and 72°C for 5 min, with a final extension step at 72°C for 7 min. 20 μl of the PCR product was the treated with 20 U of DpnI (Thermo) for 2 hours at 37°C to selectively degrade the methylated parental vector, and 20 min at 80°C to inactivate the enzyme. Then, 1μl was used to transform 50 μl of electrocompetent XL-1 E. coli cells or E. coli BL21 (DE3). Plasmids were purified using the GeneJET Plasmid Miniprep (Thermo) and the sequence verified by DNA sequencing (Macrogen).

Transformed E. coli BL21 (DE3) were grown overnight in 10 ml of Lysogeny broth (LB) containing 100 μg ml^-1 ampicillin at 37°C. For protein expression, 4 ml of overnight culture was transferred into 800 ml of LB with ampicillin and grown at 37°C to an OD₆₀₀ of 0.6-0.7. Protein synthesis was induced by adding 0.5 mM IPTG at 19°C for 24 h. Cells were then harvested by centrifugation and lysed by sonication on ice in the lysis buffer (100 mM Tris, pH 8.0, 200 mM NaCl, 10% glycerol, containing EDTA-free protease inhibitor cocktail (Amersham Biosciences). The suspension was supplemented with 1.5% Tween 20 and incubated for 90 min at 4°C with gentle agitation. Cell debris was removed by centrifugation at 15000 xg for 40 min, and the supernatant was diluted with 100 mM Tris, pH 8.0, 200 mM NaCl, and 10% glycerol to obtain a final concentration of 0.5% Tween 20. Then, imidazole was added to the supernatant to a final concentration of 10 mM and incubated with the Ni-NTA resin (GE Healthcare) equilibrated in binding buffer (100 mM Tris pH 8, 200 mM NaCl, 10% glycerol, 10 mM imidazole, and 0,5% Tween 20), for 1 hour with gentle shaking. The mixture was then packed into a column, washed with binding buffer, and the recombinant protein was eluted with elution buffer (100 mM Tris pH 8, 200 mM NaCl, 10% glycerol, 300 mM imidazole, and 0.5% Tween 20). Imidazole was rapidly extracted by using a PD-10 column equilibrated in 50 mM Tris pH 8, 100 mM NaCl, 5% glycerol, and 0.5% Tween 20. The protein was further purified by SEC using a Superdex 200 16/60 preparative grade column (GE Healthcare) previously equilibrated in SEC-buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl, 5% glycerol, 0.1% Tween 20). Fractions containing purified hTFP _α were pooled and concentrated using an Amicon Ultra-15 Millipore concentrator to 1 mg/mL. The 6xHis-tag was removed with TEV protease in a 1:20 ratio overnight at 18°C, and then the TEV was removed using a Ni-NTA resin. The identity of the recombinant hTFP_α was confirmed by MS. Two batches of recombinant protein were produced and used in the present study.

For the production of the hTFP_α-Y271F mutant, we used as a template the plasmid containing the sequence of the hTFP_α-wt, and the mutation introduced by PCR site-directed mutagenesis using the following forward and reverse primers, respectively: 5’ - GAA AAA TTG ACA GCG TTT GCC ATG ACT ATT C -3’, and 5’ - G AAT AGT CAT GGC AAA CGC TGT CAA TTT TTC -3’. The presence of the mutation was verified by DNA sequencing (Macrogen). Recombinant hTFP_α-Y271F was expressed in E. coli BL21 (DE3), following the same procedure described for the wt protein. The only difference was that the 6xHis-tag was removed with TEV protease before SEC purification, and not after this chromatography as described for the WT.

The recombinant hTFP_α-wt, hTFP_α-Y271F, and OH1-C112S produced as reported previously (Segovia et al., 2017), were phosphorylated with commercial Jak-1 kinase (Thermo Fisher Scientific) in the SEC-buffer supplemented with 2.5 mM MnCl₂, 7.5 mM MgCl₂, 1 mM DTT, and 1 mM ATP as substrate for 2 hours at 30°C. An enzyme:substrate (E:S) molar ratio of 1:3000 was used. Then, Jak-1 was removed by purification with glutathione agarose (LifeTech) and ATP was removed by SEC (PD-10, GE). The phosphorylated proteins were then used in PtpA activity and MS assays.

To evaluate PtpA activity in solution using the recombinant proteins (hTFP_α-wt or hTFP_α-Y271F, OH1-C112S) we first phosphorylated them with Jak-1 as described above. The buffer of the proteins was changed by SEC to the activity buffer (50 mM Tris-HCl pH 8.0, 50 mM NaCl, 5% glycerol, 3 mM EDTA, 1 mM DTT and 0.1% Tween 20). To avoid PtpA inactivation and interference of the detergent during MS identification of the phospho-peptides (Feist & Hummon, 2015) it was not possible to increase the concentration of detergent more than 0.1% Tween 20. To preserve both proteins in solution the maximal incubation time at 37°C was 30 min. The E:S molar ratios used were 1:5000, 1:500 and 1:50. After 30 min, spots of 5 μL of the reaction were applied in a nitrocellulose membrane by triplicates and then the membrane was dried and blocked with membrane blocking solution (Invitrogen #00-0105). Afterward, membranes were washed in TBS-T and probed for p-Tyr levels with anti-p-Tyr antibody (Cell Signalling #9411) at 1/2000 dilution in TBS-T, ON at 4°C. Blots were then incubated with horseradish peroxidase (HRP)-linked anti-mouse (Sigma-Aldrich A4416, 1/10000) secondary antibody for 1 h at RT. After four washes with TBS-T, and one wash with TBS, the reaction was developed with Pierce ECL western blotting substrate (Thermo Scientific). The chemiluminescent signals of the bands were visualized using the GBOX ChemiSystem tool (SynGene). For the assays using hTFP_α-wt or hTFP_α-Y271F as substrate, the membrane was reprobed with anti hTFP_α antibody (Abcam ab200652) to determine the ratio between p-Tyr and hTFP_α chemiluminescent signals, quantified using ImageJ (Schneider et al., 2012).

The samples were separated 1.5 cm by SDS-PAGE and stained with colloidal Coomassie Brilliant Blue G-250. Each lane was excised into 4 slices that were destained by incubation with 0.2 M ammonium bicarbonate/ACN (1:1) for 1 h at room temperature with agitation. After this, proteins were reduced with 10 mM DTT at 56°C for 60 min and then alkylated with 55 mM iodoacetamide at room temperature for 45 min and in darkness. Then, in-gel proteolytic digestion and peptide extraction were performed as described earlier (Gil et al., 2019). To verify the identity of the hTFP_α immunopurified or produced as recombinant protein and for p-Tyr identification, online MS detection/analysis was carried out in a nano-HPLC (UltiMate 3000, Thermo) coupled to a hybrid quadrupole-orbitrap mass spectrometer (QExactive Plus, Thermo). Peptide mixtures were loaded on C18 columns and separated using a two-solvent system (solvent A: 0.1% formic acid in water and solvent B: 0.1% formic acid in acetonitrile) with a gradient from 0% to 90% of B) and a flow rate of 0.2 mL/min over 100 min. Peptide analysis was performed in a Q-exactive Plus (Q-Orbitrap, Thermo) associated with a Nano HPLC. Xcalibur 2.1 was used for data acquisition of a full MS scan in the positive ion mode with m/z between 200 and 2000 m/z. Sequential fragmentation of the ten most intense ions with a normalized collision energy of 35, an isolation width of 2 m/z. The activation Q was set at 0.25, the activation time at 15 ms, and a dynamic exclusion time of 30 s. MS source parameters were set as follows: 2.3 kV electrospray voltage and 260°C capillary temperature. PatternLab V (Version 5.0.0.109) (Santos et al., 2022) was employed to generate a target-decoy database using sequences from E. coli and hTFP_α, PtpA, Jak, downloaded from the UniProt. In addition, 127 common mass spectrometry contaminants were included (Carvalho et al., 2016). The Comet search engine was operated using the following parameters: trypsin as a proteolytic enzyme with full specificity; oxidation of Met and phosphorylation on Tyr as variable modifications, carbamidomethylation of Cys as fixed modification; and 35 ppm of tolerance from the measured precursor m/z. XCorr and Z-Score were used as the primary and secondary search engine scores, respectively. Peptide spectrum matches were filtered using the Search Engine Processor (SEPro), and acceptable false discovery rate (FDR) criteria were set at ≤1% at the protein level and ≤2% at the peptide level.

To improve the identification of the phospho-peptides by MS, the phosphoproteins contained in one mg of recombinant hTFP_α were purified using the Pierce Phosphoprotein Enrichment kit (Thermo). This sample was used in the phosphatase assay (1 h at 37°C) without or with PtpA. Then the proteins were treated with sequencing-grade trypsin (0.25 μg, 3h or ON at 25°C), and the detergent was removed using a specific resin (Pirce#87780). Prior to MS analysis samples were desalted using C18 reverse phase micro-columns (Omix^®Tips, Varian), dried by vacuum, and resuspended at 1 µg/ul in 0.1% formic acid (v/v) in water. Samples were injected into a nano-HPLC (UltiMate 3000, Thermo) coupled to a hybrid quadrupole-orbitrap mass spectrometer (QExactive Plus, Thermo), and separated and analyzed as described above.