2.1. Overall Structure of PTPN12
The structure of the PTPN12 catalytic domain was solved by molecular replacement using the structure of a Lyp catalytic domain (Protein Data Bank (PDB) ID 2QCJ) as a search model and refined to 1.62 Å resolution. The overall structure of the PTPN12 catalytic domain was found to be similar to the NT4 subfamily of PTPs (PDB IDs 2OC3, 2P6X, 2QCJ, 2QCT, 3H2X, 4GFU, 4GFV, 4J51, and 4NND) [17
] with a RMSD of Cα atoms less than 1.4 Å, respectively (Figure 1
A,B). The PTPN12 catalytic domain adopted a canonical α/β folded structure typical of a central eight-stranded β-sheet, with six and two α-helices on each side.
The PTP signature motif (P-loop) was composed of residues 230–237 (Figure 1
A). The catalytic Cys231 adopted dual conformations, one of which was assigned as the phosphoryl–cysteine intermediate [23
], coupled with three water molecules (discussed below) (Figure 2
A,B). The WPD loop, which adopted a closed conformation, was well defined as residues 197–208. Other structural motifs include the pTyr recognition loop (residues 58–64) and the Q-loop (residues 278–285) [19
] (Figure 1
A). Some minor structural differences were observed between PTPN12 and other NT4 PTPs. Spatial displacements were observed in the regions around Thr81–Asp85 and Pro257–Phe260, compared with the PTPN18 and PTPN22 structures. A 5-residue insertion, which does not exist in PTPN18 and PTPN22, formed a loop between the α1 and α2 helices. In addition, residues Ser39–Lys46 of PTPN12 composed part of the α2 helix, which is similar to PTPN18, but not PTPN22 [26
2.2. Catalytic Cys231 and Water Molecules
During the early stages of the structure refinement, a clear tetrahedral-shaped density was observed at the binding site of the tyrosyl–phosphate, which was very close to the catalytic Cys231. As phosphate ions were present in the crystallization buffer, the density was assigned as a phosphate ion during the early stages of refinement. However, after several rounds of model building and refinement, the distance between the Cys231 Sγ and oxygen atom of the phosphate ion (1.3 Å) appeared to be too short. Moreover, the catalytic Cys231 adopted dual conformation. By assuming that the PTPN12 underwent a slow reaction during the protein production or crystallization procedure, Cys231 was modeled as an alternate conformation of phosphoryl–cysteine with occupancy of 0.7, and the other as an alternate conformation of unphosphorylated cysteine with occupancy of 0.3. The phosphoryl–cysteine conformation can be considered a phosphoryl–enzyme intermediate. The phosphoryl moiety interacted with the side chain of Arg237 (Arg235 in PTPN18, Arg233 in PTPN22, or Arg221 in PTP1B) and the backbone amides of the P-loop (Figure 2
A). The non-phosphorylated conformation of Cys231 adopted a different side-chain orientation with a relative rotation of ~100°.
In order to verify our hypothesis and highlight the mobility of relevant residues and structural elements, we performed a series of molecular dynamics (MD) simulations, with or without the covalent bound between the phosphate ion and the cysteine. In the first case, Cys231 was parameterized as a standard cysteine, while in the second case we derived the force field parameters of the phosphoryl–cysteine by using the method as in ref [30
]. MD simulations showed that two different orientations of the side chain of Cys231 are possible, in agreement with the two different conformations found in the crystal structure. In particular, the hydrogen atom, which is bonded to the sulfur atom of a standard cysteine side chain, slightly superimposed with the phosphate causing the reorientation of Cys321 side-chain dihedral angle by approximately 100°, similarly to what we observed in the alternate orientation of the crystal structure (Figure 3
A,B), while in the presence of the phosphoryl–cysteine we do not observe any transition (Figure 3
C). The alternative configuration, which we observe in the absence of the covalent bound between the phosphate ion and Cys231, stabilizes after the first 40 ns of dynamics, and is overall realized for approximately 90% of the corresponding MD trajectory (Figure 3
D), indicating the possibility that the two orientations of Cys231 side chains correspond to two different chemical states of the complex phosphate–cysteines. Moreover, this configuration change, which in our simulation is triggered by the protonation of Cys231 side chain, could be necessary for initiating the process of release of the phosphate ion. Considering the results of the MD simulations, an extra isolated phosphate ion was defined at the position of the phosphoryl moiety of the phosphoryl–cysteine with an optimized occupancy of 0.12 (Figure 1
B and Figure 2
A). The interactions of this phosphate ion with the surrounding residues were almost identical to those of the cysteinyl–phosphate.
Clear densities were observed for three water molecules interacting with the thiophosphate group in the crystal structure (Figure 2
B), similar to the phosphocysteine–PTP1B structure (PDB ID 1A5Y) reported previously [23
]. Among them, the W2 water molecule is ideally positioned for nucleophilic attack on the cysteinyl–phosphate intermediate, thereby forming hydrogen bonds with an oxygen atom of the cysteinyl–phosphate, the side chains of Tyr64 and Asp199, and W3 (Figure 2
A). The W1 water molecule forms hydrogen bonds with the cysteinyl–phosphate, Asp199, Gln278, Gln282, and backbone amide of the P-loop. The W3 water molecule interacts with the cysteinyl–phosphate, Asp199, Gln278, and water molecules W1 and W2 (Figure 2
As in the crystal structure, water molecules were present around the phosphate ion in the MD simulation. The simulation clearly indicated a dynamic equilibrium, wherein new water molecules replaced those that diffused away. The reorientation of the Cys231 side chain created a pocket for the water molecule (Figure 3
E). This water molecule was well positioned to push Arg237 away from the active site and thus could contribute to the release of the phosphate.
We compared the PTPN12 structure with the structures of other NT4 phosphatases. The side chain of the catalytic Cys227 residue of a substrate-free PTPN22 structure (PDB ID 2P6X) pointed to the center of the P-loop in a previous study [17
], in the same orientation as the phosphoryl–cysteine in our structure. Moreover, the catalytic cysteines in previously described phosphoryl–peptide–bound PTPN18 structures (PDB IDs 4GFU, 4GFV, and 4NND) [28
] assumed the same orientation as did the phosphoryl–cysteine in our structure. Those data indicated that Cys231 was in the same orientation immediately before and after its phosphorylation. Considering the surrounding residues, the conformation of Arg237 of PTPN12 was identical to Arg235 of phosphoryl–peptide–bound PTPN18, and Arg221 of phosphoryl–cysteine–PTP1B. In addition, a disulfide bond was observed between Cys227 and Cys129 of PTPN22 in the reduced state [27
]. It was proposed that such a disulfide bond can help stabilize the conformation of the WPD loop. In our PTPN12 structure, such a disulfide bond was not observed.
Considering all of the above observations, we suggest that the conformation of the PTPN12 catalytic domain with the phosphoryl–cysteine represented the state intermediately after the phosphate ion was transferred from the substrate to the enzyme, while the other unphosphorylated conformation was the state after the phosphate ion was released from the enzyme.
2.3. WPD Loop
The WPD loop in the PTPN12 structure was well defined as a closed conformation, which is very similar to those of PTPN18 in complex with phosphoryl–peptide (PDB IDs 4GFU, 4GFV, and 4NND) [28
]. Only Ser204 and 205 (Ser202 and 203 in PTPN18) undergo approximately 3.5 Å displacement in the comparison. Those two residues are far from the reaction center (17 Å away from the cysteinyl–phosphate). The interaction between Trp197 and Arg237 was broken in the closed WPD loop, as in previous reports [19
In the MD simulation, the position of the WPD loop fluctuated in the presence of the phosphate ion, and the loop reached a configuration (Figure 3
F, white color) that was significantly different from that of the crystal structure (Figure 3
F, green color).
During the simulation, the phosphate ion was strongly stabilized by the backbone of the residue in the P-loop. In contrast, neither the WPD loop nor the side chain of Arg237 seemed to stably interact with the ion. The former lost its interaction with the phosphate at the very beginning of the simulation (after 500 ps), while the change in its configuration created access to the ion after 30 ns and stabilized this “open” configuration. During the 200 ns simulation window, the phosphate ion was not able to diffuse spontaneously from its binding pocket, even after the observed opening of the WPD loop.
A control simulation in which the phosphate ion was removed showed lower fluctuations in the loop containing the WPD motif (0.1 nm on average, compared with 0.2 nm on average when the phosphate was present). Overall, the configuration (Figure 3
F, pink color) did not show significant differences compared with the crystal structure.
2.5. Substrate Recognition
Previously, it was reported that both PTPN18 and PTPN12 can dephosphorylate HER2 [16
]. The PTPN18/phospho-HER2 peptide structures (PDB IDs 4GFU, 4GFV, and 4NND) [28
] were superimposed onto the PTPN12 structure by superposition of the catalytic domains. In the PTPN18 structures, the pTyrs from the HER2 peptides inserted into the active site, forming hydrogen bonds and salt bridges with surrounding residues, including Asp64 (Asp66 in PTPN12) and phosphate-interacting residues in the P-loop. The phenyl rings of the pTyrs are stacked between Tyr62 and Gln276 (Tyr64 and Gln278 in PTPN12), and these crucial residues are all conserved. Specifically, HER2 L1249 can form hydrogen bonds with PTPN18 Gln276 (PTPN12 Gln278), based on the docked structure. Asp1252 of the HER2 pTyr1248 peptide forms a salt bridge with PTPN18 Arg198. That residue is substituted by His200 in the PTPN12 WPD loop. For the pTyr1196 peptide, the backbones of several C-terminal residues form hydrogen bonds with PTPN18 Arg198 (PTPN12 His200). Thus, PTPN12 His200 may play an important role in substrate selectivity (Figure 4
B). The HER2 pTyr1112 peptide adopts an extended structure, forming seven backbone hydrogen bonds with PTPN18. In the N-terminus of the pTyr1112 peptide, no side-chain atoms are involved in hydrophilic interactions. All of these conserved and non-conserved residues may reflect the substrate selectivity of PTPs in the NT4 subfamily.
2.6. Analysis of Tumor-Derived Mutations
Previous findings [16
] suggested that PTPN12 suppresses the growth and metastasis of human Triple negative breast cancer (TNBC) cells and that such function is lost by nonsynonymous mutations. Tyr1148 residue of EGFR showed the strongest differential phosphorylation in response to PTPN12 depletion. PTPN12 may also be involved in other types of tumorigenesis [32
]. We searched the publicly available Catalogue of Somatic Mutations in Cancer (COSMIC) database (http://cancer.sanger.ac.uk/cosmic
) of 29337 human primary tumors and found 133 PTPN12 mutations. Based on our crystal structure, we identified four residues in the active site of the protein, which should be crucial for the activity of PTPN12: (1) Tyr64, which interacts with the phenyl rings of substrate peptides; (2) Pro203, which is located in the WPD loop; (3) Ala233, which is located in the P-loop nearby the catalytic residue Cys231; and (4) His274, which is substituted by Met258 in PTP1B and is localized in the secondary substrate-binding pocket [31
]. We designed and generated a mutation for each of the mentioned residues (Y64C, P203S, A233T, and H274R) and tested their activity in vitro and in vivo, together with wild type PTPN12 and C231S (Figure 5
A). All purified mutants showed lower activity than the wild type PTPN12 catalytic domain in vitro (Figure 5
B). To further confirm their activities in vivo, Hela cells were transfected with the plasmids expressing wild type or mutants of the PTPN12 catalytic domain. In contrast to wild type PTPN12, the mutants showed marginal inhibition of EGFR Tyr1148 phosphorylation, even though expression of the mutants was obviously higher than the wild type protein (Figure 5
C). Since PTPN12 is a frequently inactivated tumor suppressor in several kinds of cancers [16
], our results indicate that PTPN12 inactivation by the selected mutations may play important roles in tumorigenesis in patients.