Structural basis of Irgb6 inactivation by Toxoplasma gondii through the phosphorylation of switch I

Upon infection with Toxoplasma gondii, host cells produce immune-related GTPases (IRGs) to kill the parasite. T. gondii counters this response by releasing ROP18 kinase, which inactivates IRG GTPases and inhibits their recruitment to the T. gondii parasitophorous vacuole (PV). However, the molecular mechanisms of this process are entirely unknown. Here we report the atomic structures of Irgb6 with a phosphomimetic mutation by ROP18. The mutant has lower GTPase activity and is not recruited to the PV membrane (PVM). The crystal structure shows the mutant exhibit a distinct conformation from the physiological nucleotide-free form, thus preventing GTPase cycling. This change allosterically modifies the conformation of the membrane-binding interface, preventing physiological PVM-binding. Docking simulation of PI5P also supports the impaired binding of the mutant to PVM. We thus demonstrate the structural basis for T. gondii escape from host cell-autonomous defense, and provide a structural model for regulating enzymatic activity by phosphorylation.


43
Toxoplasma gondii is an important pathogen for warm-blooded animals including humans. It 44 enters host cells and forms a membranous structure called parasitophorous vacuole (PV) in 45 order to resist host defenses [1][2]. In response to T. gondii infection, host immunity releases

77
To assess the effect of phosphorylation of the Thr95 on Irgb6, the amino acid was substituted 78 to aspartate as a phosphomimetic mutation (T95D mutation), following a previous paper [13]. 79 We confirmed that phosphorylation of Thr95 was induced by T. gondii infection in MEF cells 80 (S1 Fig). The Flag-tagged wild-type or T95D mutant of Irgb6 (Irgb6-T95D) was expressed in 81 Irgb6-deficient MEFs (Fig 1A). An indirect immunofluorescence study showed that Flag-82 tagged wild-type Irgb6 (Irgb6-WT) was loaded onto T. gondii PVM (Fig 1A). In sharp 83 contrast, Flag-tagged Irgb6-T95D was not detected on T. gondii PVM at all (Fig 1A), 84 suggesting that the phosphorylation of Thr95 of Irgb6 is required for the localization on the 85 PVM. Next, we examined the effect of Thr95 phosphorylation on recruitment of other IFN-86 inducible GTPases (Fig 1B). The recruitment of Irga6 and Irgb10 was not observed upon 87 reconstitution of Irgb6-T95D in Irgb6-deficient MEFs ( Fig 1B). IFN-γ stimulates coating of 88 effectors such as ubiquitin and p62/Sqstm1 on T. gondii PVM in a manner dependent on IFN-89 inducible GTPases [15][16]. Therefore, we next analyzed whether loading of ubiquitin and 90 p62/Sqstm1 is affected by the Thr95 phosphorylation ( Fig 1C). Although reconstitution of 91 Irgb6-WT recovered effector loading on T. gondii PVMs in Irgb6-deficient MEFs, 92 reconstitution of Irgb6-T95D did not (Fig 1C). Taken together, these data indicate that Thr95  Irgb6-T95D mutation disrupts GTPase activity. 97 We then examined the GTPase activity of Irgb6-T95D purified through anion exchange 98 chromatography and compared the result to that of Irgb6-WT. Size exclusion chromatography 99 (SEC) of purified Irgb6-T95D produced two peaks, similar to Irgb6-WT (Fig 2A). SDS-PAGE 100 analysis indicated that Irgb6-T95D was mainly eluted in the second peak, which was used for 101 in vitro functional and structural assays.   The crystal structures of Irgb6-T95D were solved with GTP (Irgb6-T95D-GTP) and without 123 nucleotide (nucleotide free: NF) (Irgb6-T95D-NF) at 1.68 and 2.05 Å resolution, respectively 124 ( Fig 3A and 3B and Table 1). The former structure possessed GTP in the nucleotide binding 125 pocket ( Fig 4A). The latter had neither nucleotide nor ion in the pocket (Fig 4B and 4C).

127
Irgb6-T95D-GTP adopted the same conformation as Irgb6-WT-GTP (Fig 3A and 3C), with an 128 overall root mean square deviation (RMSD) from Irgb6-WT-GTP of 0.259 Å ( Fig 3F). As in 129 the Irgb6-WT-GTP structure [12], Irgb6-T95D-GTP represents a collision complex with GTP 130 in which the switch I and II regions are located away from the γ-phosphate of GTP ( Fig 4A).   Irgb6-T95D adapts an atypical Apo-form. 152 As described above, the structure of the G-domain of Irgb6-T95D-NF differed significantly 153 from that of Irgb6-WT-NF. These differences are readily visible in the active site of the G-154 domain with the electron density map (Fig 4). It should be noted that the resolution of both side of H5 as the pivot point (Fig 5B and S1 Movie). In this process, the rotation angle of helix αA is relatively more extensive than that of the other helix movements so that the groove 176 between αA and αF is widened, along with the flipping of Trp3 (Figs 5B and 5C and S3). Due 177 to these changes, the αK-αLa loop can be inserted into the widened groove, further stabilizing 178 the elongated form of helix αLa through the hydrophobic contact between Trp3 and Val364.

180
Our previous reports indicate that the amino acid triangle formed by Trp3, Lys275,and 181 Arg371, is essential for PVM binding [11][12]. Quite suggestively, we found that Trp3 and   As previously reported, the head group of PI5P bound to the region surrounded by Trp3, 201 Lys275, and Arg371 in Irgb6-WT-GTP. The tips of the phosphate groups of PI5P were generally oriented to bind to Arg371 or Lys 275 [ Fig 6A]. The tails of the head group 203 generally pointed to the opposite side, implying that the acyl chain would extend toward the 204 N-terminal helices. Another characteristic of Irgb6-WT-GTP is that the PI5P binding site is 205 relatively broad. In the case of Irgb6-T95D-NF, on the other hand, PI5P was found to bind to a 206 completely different location outside of the original binding site (Fig 6B and 6C). This region 207 is surrounded by R371 and Trp3, and thus this unusual pose was assumed to be due to the   This study found that the phosphomimetic mutation induced structural changes in the 223 nucleotide-free state rather than to the GTP collision complex. This means that Irgb6 cannot 224 take on the active form (which proceeds to GTP hydrolysis just after GTP binding) unless 225 Irgb6 assumes the physiologically relevant nucleotide-free form. We recently reported that 226 Irgb6 biochemically binds to the membrane in a nucleotide-free form and distorts the 227 membrane through GTP binding [18]. It is consistent with the present finding that phosphorylation by ROP18 prevents Irgb6 from taking on the physiological nucleotide-free 229 form and also changes the membrane-binding interface to inhibit membrane binding.

231
The conformational change in Irgb6 is transmitted to helices H5 and αE from the G-domain 232 and further converted into rotational movement of the N-and C-domains (Fig 5). Phospholipid

254
Immunofluorescence assay 255 MEFs were plated at equal densities (1.5 x 10 5 cells per well in a 6-well plate) on glass coverslips and 256 exposed to 10 ng/ml IFN-γ for 18-20 h at 37 ℃. The cells were infected with ME49 tachyzoites at 257 MOI 4 and incubated at 37 ℃ for 2 h. Next, cells were fixed in PBS containing 3.7% 258 paraformaldehyde for 10 min. Cells were then permeabilized with 0.002% digitonin in PBS for 10 min 259 and blocked with 8 % FBS in PBS for 10 min.  Data collection and structure determination 335 Single crystals were mounted in LithoLoops (Protein Wave) with the mother liquor containing 10%

336
(v/v) glycerol as a cryoprotectant and were frozen directly in liquid nitrogen before X-ray experiments.

337
Diffraction data collection was performed on the BL32XU beamline at SPring-8 using the automatic 338 data collection system ZOO (19). The diffraction data were processed and scaled using the automatic 339 data processing pipeline KAMO (20). The structure was determined using PHENIX software suite 340 (21). Initial phase was solved by molecular replacement using the Irgb6-WT crystal models (PDB ID: 341 7VES and 7VEX) with phenix.phaser. The initial model was automatically constructed with 342 phenix.AutoBuild. The model was manually built with Coot (22) and refined with phenix.refine. The 343 statistics of the data collection and the structure refinement are summarized in Table 1. UCSF Chimera 344 (23) was used to create images and compare structures. protein grids were 20 × 20 × 20 Å in size. Eight grids were generated for the 8 structures.

358
Ligand Preparation: The Pi5P molecule was truncated up to the polar head before being prepared.

359
LigPrep from Schrödinger suite 3 was used to produce low-energy, three dimensional (3D) ligands with 360 correct chirality. Three conformations were generated and used for docking.