Structural insights and in vitro reconstitution of membrane targeting and activation of human PI4KB by the ACBD3 protein

Phosphatidylinositol 4-kinase beta (PI4KB) is one of four human PI4K enzymes that generate phosphatidylinositol 4-phosphate (PI4P), a minor but essential regulatory lipid found in all eukaryotic cells. To convert their lipid substrates, PI4Ks must be recruited to the correct membrane compartment. PI4KB is critical for the maintenance of the Golgi and trans Golgi network (TGN) PI4P pools, however, the actual targeting mechanism of PI4KB to the Golgi and TGN membranes is unknown. Here, we present an NMR structure of the complex of PI4KB and its interacting partner, Golgi adaptor protein acyl-coenzyme A binding domain containing protein 3 (ACBD3). We show that ACBD3 is capable of recruiting PI4KB to membranes both in vitro and in vivo, and that membrane recruitment of PI4KB by ACBD3 increases its enzymatic activity and that the ACBD3:PI4KB complex formation is essential for proper function of the Golgi.


Supplementary Discussion
The TBC1 domain family members 22A (TBC1D22A) and 22B (TBC1D22B) were recently shown to interact with the Q domain of ACBD3 1 . Members of this family were proposed to act as GTPase activating proteins for Rab family members which are involved in vesicle trafficking.
TBC1D22A and B share a high degree of homology, are Golgi membrane localized, and act as putative RabGAPs for Rab33. Given the similarity of the primary sequences of the ACBD3 binding sites of PI4KB and TBC1D22A/B (SI Fig. 8A), we were able to generate homology models of the ACBD3-TBC1D22A and ACBD3-TBC1D22B complexes (SI Fig. 8B). These models are in a good agreement with previously published observations that the ACBD3-PI4KB and ACBD3-TBC1D22A/B interactions are mutually exclusive, suggesting a possible regulatory mechanism for Golgi recruitment of the PI4KB enzyme 1 .

SPR (Surface plasmon resonance) and AUC (Analytical ultracentrifugation) -SPR measurements were
performed on a four-channel SPR sensor platform (PLASMON IV) developed at the Institute of Photonics and Electronics, AS CR, Prague. A gold chip was first functionalized with alkanethiols containing carboxylic terminal groups (Prochimia) in pure ethanol, and then mounted on the prism of an SPR sensor. All experiments were performed at 25 °C at a flow rate of 30 µL/min. Activation of carboxylic terminal groups on the sensor surface was performed in situ with a solution of Nhydroxysuccinimide and N-ethyl-N-(dimethylaminopropyl)-carbodiimide hydrochloride (Biacore). Then, a 0.02 mg/mL neutravidin solution in the SA buffer (10 mM sodium acetate pH 5) was loaded, followed by a high ionic strength solution (10 mM sodium phosphate pH 7.4, 0.5 M NaCl) to wash out noncovalently bound neutravidin, and 1 M ethanolamine to deactivate residual carboxylic groups. For immobilization a 100 nM solution of an appropriate recombinant biotinylated protein in the running buffer (10 mM Tris pH 8, 200 mM NaCl, and 1 mM TCEP) was used to achieve approximately 2 nm change in the relative response signal (the protein biotinylation was achieved by co-expression of the appropriate protein tagged with an AviTag at its N-terminus with a biotin ligase BirA in E. coli as described in Kay et al. 2 ). Afterwards, untagged proteins in the indicated series of concentrations in the running buffer were injected for 3 min and then the dissociation was monitored for another 5 min. The data were fit to a single-exponential model. Rate constants of association and dissociation were obtained by fitting the observed change in resonance signal using the following equations: (1) (2) where c is the protein concentration, t is time, k on is the association rate constant, k off is the dissociation rate constant, D 1 and D 2 are the linear drift terms, and R as , R dis , R 0 , R 1 , and R max are corresponding changes in the relative resonance signal.
NMR spectroscopy -The families of converged structures for the ACBD3 free Q domain and the ACBD3:PI4KB complex was initially calculated using Cyana 2.1 3 . The combined automated NOE assignment and structure determination protocol was used to automatically assign the NOE cross-peaks identified in NOESY spectra and to produce preliminary structures. In addition, backbone torsion angle constraints, generated from assigned chemical shifts using the program TALOS+ 4 were included in the calculations. Subsequently, five cycles of simulated annealing combined with redundant dihedral angle constraints were used to produce sets of converged structures with no significant restraint violations (distance and van der Waals violations <0.2Å and dihedral angle constraint violation <5°), which were further refined in explicit solvent using the YASARA software with the YASARA forcefield 5 . The structures with the lowest total energy were selected. Analysis of the family of structures obtained was carried out using the Protein Structure Validation Software suite (www.nesg.org) and Molmol 6 . The statistics for the resulting structures are summarized in SI Table 1.
Giant Unilamellar Vesicle Preparation and Imaging -Giant Unilamellar Vesicles (GUVs) of the desired composition were prepared by electroformation. 50 µg of the lipid mixture was applied on each electrode (5 x 5 cm ITO coated glass) and dried in vacuum overnight. The next day the coated glasses were moved to a home-made teflon chamber and 5 mL of 600 mM sucrose heated to 60 °C was added. Altering current with a maximum amplitude of 1V and frequency of 10 Hz was applied for 1 hour while keeping the chamber at 60 °C. For imaging 100 µL of GUVs and 100 µL of buffer (50 mM Tris pH = 8, 300 mM NaCl, 1 mg/mL BSA) containing appropriate proteins were mixed. The ATTO647N-DOPE and Alexa488 or CFP (mCerulean) labeled proteins were excited simultaneously by 640 nm and 488 nm or 405 nm lasers and imaged using a Zeiss LSM780 confocal microscope.

Video legends
1) Video 1: Mitochondria recruitment experimentwt Q domain. Cells transfected with AKAP1-FRB-CFP, GFP-PI4KB and wild-type Q domain-FKBP-mRFP constructs filmed during addition of rapamycin. AKAP1-FRB-CFP is localized on the mitochondria. Note that when rapamycin is added (time 1 min 15 s) the wild-type Q domain-FKBP-mRFP rapidly translocates to mitochondria as well and is followed with ~ 2 min delay by the GFP-PI4KB. 2) Video 2:Mitochondria recruitment experiment -H 264 A Q domain. The same experiment as in video 1 performed using the H 264 A Q domain mutant. Note that the mutant Q domain translocates to the mitochondria as well but is not able to recruit the GFP-PI4KB.

Supplementary Tables
SI Table 1