HDX‐MS experiments were carried out on the full‐length construct of PI4KIIIβ in the presence of the GTPase Rab11. HDX‐MS is dependent on the generation of an optimized peptide map for localization of amide exchange rates throughout the protein. The optimized map for PI4KIIIβ is shown in Figure 1, and was composed of 125 individual peptic peptides, which covered 88% of the amide hydrogens of PI4KIIIβ.

The initial goal of HDX‐MS experiments was to define dynamic regions (either no or very transient secondary structure) in PI4KIIIβ that might be removed for crystallographic analysis. Initial HDX experiments were carried out using a very short pulse of deuterium (3 s of D₂O exposure at 0°C) so that only amides with a very low content of secondary structure would be deuterated. The exchange profile of PI4KIIIβ is shown in Figure 2 mapped onto the sequence of PI4KIIIβ. There were a number of regions within PI4KIIIβ that showed very high levels of deuterium incorporation (>90% deuterium incorporation) even at these very short exposure times. Recent evidence suggests that the primary, but not sole, determinant of the rate of exchange of amide hydrogens is the involvement in secondary structure,8, 9, 27 suggesting that these rapidly exchanging regions in PI4KIIIβ are either intrinsically disordered or contain very dynamic secondary structure elements. These regions included the N‐terminus (residues 1–130), a serine rich region from residues 242 to 289, a long linker from residues 408–507 located within the kinase domain, the activation loop in the kinase domain, and the C‐terminal residues 785–801.

We also wanted to establish the important regions in both PI4KIIIβ and Rab11 that mediated the contacts between this complex. We carried out HDX‐MS experiments on both Rab11 and PI4KIIIβ at a number of time points (3, 30, 300 s at 21°C, and 3 s at 0°C), and significant decreases (>6% and 0.6 Da change in deuterium exchange at any time point, paired t‐test P < 0.01) were seen in peptides covering the helical region of PI4KIIIβ, and throughout numerous regions within Rab11 (Fig. 3).

Using the HDX information describing both dynamic regions and conformational changes accompanying Rab11 binding, we designed a number of deletions [Fig. 4(A)] including an N‐terminal deletion from 1 to 120, two internal deletions (one from 249 to 287, and one from 408 to 507), and a C‐terminal deletion from 785 to 801. Individual deletions were screened for their expression, with each expressing at a similar level to the wild type enzyme. All deletions were then combined into a single construct that contained two internal deletions, as well as removing the N‐terminus and the C‐terminus. This construct, referred to from now on as xtalPI4K (HsPI4KB S294A 121–784, Δ249–287, Δ408–507), enabled us to generate crystals of PI4KIIIβ in complex with Rab11. Diffracting crystals were only obtained in the presence of the Rab11 binding partner [Fig. 4(A)].

A major concern when engineering proteins for structural studies is that the introduction of truncations may disrupt the native conformation, as well as the function of the enzyme. To verify that the HDX optimized construct was not structurally perturbed by the engineered deletions we carried out tests on both the structure and function of this enzyme compared with the full‐length enzyme. HDX‐MS experiments were carried out on both full length PI4KIIIβ and xtalPI4K, and these revealed that levels of deuterium incorporation for xtalPI4K were similar to full length PI4KIIIβ [Fig. 4(B)]. Importantly, there was no difference in dynamics within or around the active site of the enzyme. To verify that truncations did not change the activity or ability to bind protein partners, we carried out lipid kinase assays as well as GST pulldowns with GST‐tagged Rab11. We have previously shown that the C‐termini of both PI4K and PI3Ks are essential for their activity.28, 29 For this reason, we made a construct containing the N‐terminal and internal deletions, but with an intact C‐terminus (referred to as xtal PI4KIIIβ +cterm). This construct showed a slightly reduced (∼70%) lipid kinase activity on pure PI vesicles compared to WT full length PI4KIIIβ [Fig. 4(D)]. Binding assays were carried out with the PI4KIIIβ binding partner Rab11 [Fig. 4(C)], and GST pull downs indicated that there was no qualitative difference in the ability of the xtalPI4K to bind to Rab11 compared with full length PI4KIIIβ.

Using the HDX optimized constructs of PI4K and Rab11 we solved the structures of PI4KIIIβ in complex with Rab11‐GTPγS (2.65 Å) and in complex with both Rab11‐GDP as well as the potent small molecule inhibitor BQR695 (3.2 Å, Crystallographic details in Table 1). The HDX levels are shown mapped onto the structure of PI4KIIIβ [Fig. 5(A)]. The full details of the interaction of PI4KIIIβ with Rab11 were described in a recent manuscript that solved the structure of an intermediate deletion construct of PI4KIIIβ (121–784 with an internal deletion from 408 to 507) bound to Rab11 at 2.99 Å resolution.28 This structure revealed the molecular basis of its interaction with Rab11, as well as the molecular details of its interaction with the potent small molecule inhibitor PIK‐93, however, this construct was refractive to the generation of higher resolution crystals and only crystallized in the presence of PIK‐93.

It has been postulated previously that PI4KIIIβ can bind to both GDP and GTP loaded Rab11,28 with a slight preference for the GTP loaded state. The structure of GDP loaded Rab11 bound to PI4KIIIβ reveals that the interface between these proteins is similar, to the interface for GTP loaded Rab11 bound to PI4KIIIβ, with the switch regions in Rab11 being disordered when bound to GDP [Fig. 5(B,C)], similar to what has been reported for previously reported GDP loaded Rab11 protein complexes.30

The higher resolution crystals of the PI4KIIIβ bound to GTPγS‐Rab11 revealed novel molecular details of the PI4KIIIβ‐Rab11 interface. These crystals revealed formation of a helical element (residues 70–77) within switch 2 (residues 72–82) of Rab11 that was not present in structures of Rab11 GTPγS alone.31 This region of Rab11 is located at a crystallographic contact site with a symmetry‐related PI4KIIIβ molecule. To examine if this conformational change occurs in solution, we examined the H/D exchange information of Rab11 in the presence and absence of PI4KIIIβ, mapped on to the structure. Notably, there was a decrease in exchange in the region from 72 to 80 within switch 2 of Rab11 [Fig. 6(A)], and there is no direct contact between this region and PI4KIIIβ. This confirmed that there is indeed a conformational change in solution of the switch regions of Rab11 in the presence of PI4KIIIβ.

The major consequence of the changed orientation of switch 2 is a reorientation of the hydrophobic triad of Rab11 (Phe48, Trp65, and Tyr80), which is involved in binding to downstream effector proteins [Fig. 6(B–D] (Supporting Information movie 1).

The fully optimized xtalPI4KIIIβ construct in complex with Rab11 formed crystals in the absence of small molecule inhibitors, and crystallization was amenable to soaking with a variety of small molecule inhibitors. Examination of the active site of Apo PI4KIIIβ (from the GTP revealed a small conformational change compared to PI4KIIIβ bound to PIK93, with K549 and D674 forming a salt bridge that occupies the active site [Fig. 6(A)].

The structure of PI4KIIIβ bound to the potent anti‐malarial compound BQR695 in complex with GDP loaded Rab11 was refined to a resolution of 3.2 Å [Fig. 7(B,C)]. The compound BQR695 is a recently reported highly potent PI4KIIIβ inhibitor that has high potency against both the human (IC50 = 80 nM) and Plasmodium vivax (∼3.5 nM) variants of PI4KIIIβ.26 The binding mode of this compound was unambiguous (Supporting Information Fig. S1), and the residues mediating this interaction are shown in Figure 7. This compound makes two putative hydrogen bonds with PI4KIIIβ, one between the nitrogen of the central quinoxaline and the amide hydrogen of V598, and one between the hydrogen on the amino group off the central quinoxaline with the carbonyl of A601. The hydrogen bond between V598 and the central quinoxaline is characteristic of many protein and lipid kinase inhibitors,32 as this mimics the hydrogen bond made with the N1 of the ATP adenine. This compound has been shown to be specific over PI3K kinases (10 fold more potent against PI4KIIIβ relative to p110α, and >100 fold more potent against PI4KIIIβ compared with all other class I and class III PI3K isoforms). Comparing the structure of PI4KIIIβ bound to BQR695 compared with structures of different PI3K inhibitor structures33, 34, 35, 36 reveals the molecular basis for this specificity. PI4KIIIβ has a larger binding pocket to accommodate the glycyl methyl amide group compared to the class I and class III PI3Ks. The wall of this pocket is composed of L383 in PI4KIIIβ, which corresponds to a tryptophan in class I PI3Ks, and a phenylalanine in class III PI3Ks [Fig. 7(D,E)]. The clash at this position is likely to interfere with the ability of the central quinoxaline of BQR695 to properly hydrogen bond with V598 of PI4KIIIβ.

Truncated human PI4KIIIβ and full‐length human Rab11a (Q70L) were expressed in BL21 C41 (DE3) cells. For Rab11a (Q70L) expression, cultures were grown to an OD₆₀₀ of 0.7 and induced with 0.5 mM IPTG for 3.5 h at 37°C. For truncated PI4KIIIβ expression, cultures were induced overnight at 16°C with 0.1 mM IPTG at an OD₆₀₀ of 0.6. Cells were harvested by centrifugation, washed with cold phosphate‐buffered saline (PBS), frozen in liquid nitrogen, and pellets were stored at −80°C. Full length PI4KIIIβ was expressed from Spodoptera frugiperda (Sf9) cells by infecting 1–4 L of cells at a density of 1.0 × 10⁶ cells/mL with baculovirus encoding the kinase. All PI4K constructs had an N‐terminal 6xhis‐tag followed by a TEV protease site. After 48–65 h infection at 27°C, Sf9 cells were harvested and washed with ice‐cold PBS.

Pellets of cells expressing PI4KIIIβ were resuspended in lysis buffer [20 mM Tris‐HCl pH 8.0 (4°C), 100 mM NaCl, 10 mM imidazole, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol, protease inhibitor cocktail (Millipore Protease Inhibitor Cocktail Set III, Animal‐Free)], and were sonicated on ice for 5 min. Triton X‐100 was then added to a final concentration of 0.2%, and the lysate was centrifuged for 45 min at 20,000g. The supernatant was then filtered through a 0.45 µm filter (Celltreat Scientific Products) and was loaded onto a 5 mL HisTrap FF column (GE Healthcare) equilibrated in buffer A [20 mM Tris‐HCl pH 8.0 (4°C), 100 mM NaCl, 10 mM imidazole, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol]. The column was washed with 20 mL of buffer A, followed by 20 mL of 6% buffer B (20 mM Tris‐HCl pH 8.0, 100 mM NaCl, 200 mM imidazole, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol), and was eluted with 100% buffer B. The His‐affinity tagged protein was cleaved overnight at 4°C with TEV protease. The cleaved protein was then diluted to 50 mM NaCl (using 20 mM Tris‐HCl pH 8.0, 10 mM imidazole, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol) and was loaded onto a 5 mL HiTrap Q HP column (GE Healthcare) equilibrated in buffer C (20 mM Tris‐HCl pH 8.0, 50 mM NaCl, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol). Protein was eluted with a gradient elution using buffer D (20 mM Tris‐HCl pH 8.0, 1.0M NaCl, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol). Fractions containing the cleaved PI4KIIIβ were pooled and concentrated to 700 μL in an Amicon 50 K centrifugal filter (Millipore). The protein was then loaded onto a HiPrep 16/60 Sephacryl S200 column equilibrated in buffer E [20 mM HEPES pH 7.2, 150 mM NaCl, 1 mM Tris Carboxyl Ethyl Phosphine (TCEP)]. The cleaved PI4KIIIβ was then concentrated to <15 mg/mL in an Amicon 50 K centrifugal filter (Millipore), and aliquots were frozen in liquid nitrogen and stored at −80°C.

Pellets of cells expressing Rab11 were resuspended in lysis buffer lacking imidazole [20 mM Tris‐HCl pH 8.0, 100 mM NaCl, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol, protease inhibitor cocktail (Millipore Protease Inhibitor Cocktail Set III, Animal‐Free)]. Cells were sonicated and centrifuged as described for PI4KIIIβ. The supernatant was filtered through a 0.45 µm filter (Celltreat Scientific Products) and incubated for 1 hour with 4 mL of Glutathione Sepharose 4B beads (GE Healthcare) equilibrated in buffer F (20 mM Tris‐HCl pH 8.0, 100 mM NaCl, 5% (v/v) glycerol, 2 mM β‐mercaptoethanol) followed by a 3 × 15 mL wash in buffer F. The GST tag was cleaved overnight on the beads with TEV protease. Anion‐exchange chromatography was performed as outlined above for the truncated PI4KIIIβ. Cleaved and GST‐tagged Rab11a(Q70L) were then concentrated to between 5 and 15 mg/mL and nucleotide loaded by adding EDTA to 10 mM followed by 1 U of phosphatase (Phosphatase, Alkaline‐Agarose from calf intestine, Sigma P0762‐100UN) per mg of protein. Proteins were then incubated for 1.5 h. The phosphatase was removed using a 0.2 µm spin filter (Millipore); the flow‐through was collected, and a 10‐fold molar excess of GTPγS or GDP was added followed by MgCl₂ to a final concentration of 20 mM. Proteins were incubated for 30 min. GST‐tagged Rab11a(Q70L) was aliquoted and frozen in liquid nitrogen. Gel filtration was performed with cleaved GDP or GTPγS‐loaded Rab11a(Q70L) as described above for PI4KIIIβ.

Glutathione Sepharose 4B beads (GE Healthcare) were washed three times by centrifugation and resuspension in fresh buffer G (20 mM Hepes pH 7.0, 100 mM NaCl, 2 mM TCEP) at 4°C. GST‐tagged bait protein [GST control or GST‐Rab11a Q70L (GTPγS)] was then added to a concentration of 5 μM and incubated with the beads on ice for 30 min. Beads were then washed three times with buffer G at 4°C. Nontagged prey proteins (PI4KIIIβ or truncated PI4KIIIβ) were then added to a final concentration of 5 μM at which point the input was taken for SDS PAGE analysis. The mixture was incubated on ice for an additional 30 min and then washed four times with buffer G at 4°C, at which time an aliquot was taken for SDS PAGE analysis.

One hundred nanometer extruded PI vesicles were made with soybean phosphatidylinositol (Sigma) in lipid buffer [20 mM HEPES pH 7.5 (RT), 100 mM KCl, 0.5 mM EDTA] using the Avanti lipid mini‐extruder. Lipid kinase assays were carried out using the Transcreener® ADP² FI Assay (BellBrook Labs) following the published protocol as previously described44; 4 µL Reactions ran at 21°C for 30 min in a buffer containing 30 mM Hepes pH 7.5 (RT), 100 mM NaCl, 50 mM KCl, 5 mM MgCl₂, 0.25 mM EDTA, 0.4% v/v Triton‐X, 1 mM TCEP, 0.5 mg/mL PI vesicles and 10 μM ATP. Both full length human PI4KIIIB and the truncated PI4KIIIB crystal construct with the c‐term were run at 200 nM. Fluorescence intensity was measured using a Spectramax M5 plate reader with λex = 590 nm and λem = 620 nm (20‐nm bandwidth).

HDX reactions were conducted with 10 μL of protein in Dilution Buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM TCEP), and initiated by the addition of 40 μL of D₂O Buffer Solution (10 mM HEPES pH 7.5, 50 mM NaCl, 2 mM TCEP, 92% D₂O), to give a final concentration of 74% D₂O. Final protein concentrations were 1 μM. A fully deuterated sample was generated by incubating protein with 1M guanidine‐HCl for 30 min, followed by overnight incubation with deuterated buffer at a final concentration of 74% D₂O. Hydrogen exchange was terminated by the addition of a quench buffer (final concentration 0.6 M guanidine‐HCl, 0.8% formic acid). Samples were rapidly frozen in liquid nitrogen and stored at −80°C until mass analysis.

Protein samples were rapidly thawed and injected onto a UPLC system immersed in ice as previously described.45 The protein was run over an immobilized pepsin column (Applied Biosystems; porosyme, 2‐3131‐00) at 130 μL/min for 3 min and the peptides were collected onto a VanGuard precolumn trap (Waters). The trap was subsequently eluted in line with an Acquity 1.7 μm particle, 100 × 1 mm² C18 UPLC column (Waters), using a gradient of 5‐36% B (buffer A 0.1% formic acid, buffer B 100% acetonitrile) over 20 min. Mass spectrometry experiments were performed on both a Xevo QTOF (Waters) as well as an Impact II TOF (Bruker) acquiring over a mass range from 350 to 1500 m/z for 30 min, using an electrospray ionization source operated at a temperature of 225°C, and a spray voltage of 2.5 kV (Xevo) or a temperature of 200°C, and a spray voltage of 4.5 kV (Impact).

Peptide identification was done by running tandem MS/MS experiments using a 5–36% B gradient over 120 min. This was supplemented with a 20 min MS/MS gradient separation to identify and correct the retention time for all samples. MS/MS was run in data dependent acquisition mode with a 1 s precursor scan from 350 to 1500 m/z, followed by three fragment scans from 50 to 2000 m/z of 2s (Xevo) or a 0.5 s precursor scan from 200‐2000 m/z, followed by 12 fragment scans from 150 to 2000 m/z of 0.25 s (Impact). The resulting MS/MS datasets were analyzed with the Mascot search within Mascot distiller (Matrix Science). The MS tolerance in mascot was set to 3 ppm with an MS/MS tolerance at 0.1 Da. All peptides with a Mascot score >20 were analyzed using HD‐Examiner Software (Sierra Analytics). The full list of peptides was then manually validated by searching a non‐deuterated protein sample's MS scan to test for the correct m/z state, and check for the presence of overlapping peptides. Ambiguously identified peptides were excluded from all subsequent analysis. Marker peptides identified in the 20 and 120 min samples were used to adjust the retention time of the long MS/MS gradient before analysis of deuterium exchange. Retention time adjustment was carried out using HD‐Examiner Software (Sierra Analytics). The first round of analysis and identification were performed automatically by the HD‐Examiner software, but all peptides (deuterated and non‐deuterated) were manually verified at every state and time point for the correct charge state, m/z range, presence of overlapping peptides, and any deviation from the expected retention time. Any peptide that deviated from the expected mass by greater than 5 ppm was excluded from analysis. Corrections for back exchange were generated from a fully deuterated sample. All experiments were carried out in triplicate.

Crystals of the optimized PI4KIIIβ construct with Rab11a‐GTPγS or Rab11a‐GDP were obtained in vapor phase equilibration plates in sitting drops. The reservoir solution was 15% (w/v) PEG‐4000, 100 mM sodium citrate pH 5.6, 200 mM ammonium sulfate and three volumes of protein were mixed with one volume in the crystallization drops, with a final drop volume of 2.5 μL. Refinement plates were set by gridding PEG‐4000, ammonium sulfate, and glycerol. Optimized crystals were obtained by seeding using the Hampton Research Seed Bead Kit according to the manufacturer's instructions. The best crystals were obtained in 13–15% (w/v) PEG‐4000, 100 mM sodium citrate pH 5.6, 250 mM ammonium sulfate, 2% glycerol with a 1/1000 or 1/10,000 seed solution dilution, a Rab11a‐GTPγS final concentration of 4.51 mg/mL and a PI4K final concentration of 7.38 mg/mL. Crystals were frozen in liquid nitrogen using cryo buffer [15% PEG‐4000 (w/v), 100 mM sodium citrate pH 5.6, 250 mM ammonium sulfate, 25% (v/v) glycerol] cryoprotectant.

Inhibitor soaks were performed by incubating crystals with 0.5 μL of 10 μM inhibitor stocks in cryo buffer for 30 min, followed by a 30 min incubation with 0.5 μL of 100 μM inhibitor stock in cryo buffer, and a final 30 min incubation in 1 mM inhibitor stock in cryo buffer. Before the final addition, 1 μL was removed from the crystal drop and 1 μL of the 1 mM inhibitor in cryo buffer was added.

Data were collected at 100 K at the Canadian Macromolecular Crystallography Facility (Canadian Light Source, CLS) beamline 08ID‐1. Data were integrated using iMosflm 7.1.146 and scaled with AIMLESS.47 Phases were initially obtained by molecular replacement using Phaser,48 with the structure of PI4KIIIβ bound to PIK‐93 and Rab11 (pdb code: 4D0L) used as the search model. The final model of Apo PI4KIIIβ bound to Rab11 was built using iterative model building in COOT49 and refinement using Phenix50, 51 to R _work = 21.6 and R _free = 24.6. The final model of PI4KIIIβ bound to BQR695 in complex with GDP loaded Rab11 was refined to R _work =25.5 and R _free = 28.6. The binding mode of BQR695 was unambiguous, and ligand geometry was generated using the elbow subset of Phenix.52 Full crystallographic statistics are shown in Table 1. Electron density corresponding to GTPγS and GDP in Rab11 was clear and the mode of binding was unambiguous.

The BQR695 compound was synthesized according to the published protocol in McNamara et al.26