IRBIT Interacts with the Catalytic Core of Phosphatidylinositol Phosphate Kinase Type Iα and IIα through Conserved Catalytic Aspartate Residues

Phosphatidylinositol phosphate kinases (PIPKs) are lipid kinases that generate phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2), a critical lipid signaling molecule that regulates diverse cellular functions, including the activities of membrane channels and transporters. IRBIT (IP3 R-binding protein released with inositol 1,4,5-trisphosphate) is a multifunctional protein that regulates diverse target proteins. Here, we report that IRBIT forms signaling complexes with members of the PIPK family. IRBIT bound to all PIPK isoforms in heterologous expression systems and specifically interacted with PIPK type Iα (PIPKIα) and type IIα (PIPKIIα) in mouse cerebellum. Site-directed mutagenesis revealed that two conserved catalytic aspartate residues of PIPKIα and PIPKIIα are involved in the interaction with IRBIT. Furthermore, phosphatidylinositol 4-phosphate, Mg2+, and/or ATP interfered with the interaction, suggesting that IRBIT interacts with catalytic cores of PIPKs. Mutations of phosphorylation sites in the serine-rich region of IRBIT affected the selectivity of its interaction with PIPKIα and PIPKIIα. The structural flexibility of the serine-rich region, located in the intrinsically disordered protein region, is assumed to underlie the mechanism of this interaction. Furthermore, in vitro binding experiments and immunocytochemistry suggest that IRBIT and PIPKIα interact with the Na+/HCO3 − cotransporter NBCe1-B. These results suggest that IRBIT forms signaling complexes with PIPKIα and NBCe1-B, whose activity is regulated by PI(4,5)P2.

To gain further insight into the function of IRBIT, we previously performed proteomic analysis of IRBIT-interacting proteins [20]. One of these proteins belongs to the PIPK family. As described above, PI(4,5)P 2 produced by PIPKs functions as a precursor of IP 3 , an agonist of IP 3 R. In addition, PI(4,5)P 2 regulates the activity of NBCe1-B [13,35]. These relationships between PI (4,5)P 2 and IRBIT-binding proteins suggest that IRBIT is involved in forming signaling complexes containing PIPKs and PI(4,5)P 2 effectors. In this study, we analyzed the interactions of IRBIT with the PIPK family. We determined the isoform specificity of the interaction between IRBIT and PIPK family members. We elucidated the intriguing molecular mechanisms by which the phosphorylation sites of IRBIT bind to the catalytic core of PIPKs. Furthermore, we described potential signaling complexes containing IRBIT, PIPKs, and NBCe1-B.

Plasmids
Hemagglutinin (HA)-tagged IRBIT and its site-directed mutants were described previously [3]. Deletion mutants of IRBIT were generated by subcloning truncated cDNA fragments of IRBIT into pHM6 (Boehringer Mannheim). pcDNA3-IRBIT was generated by subcloning the cDNA encoding mouse IRBIT into pcDNA3 (Invitrogen). Mouse PIPKIα, PIPKIβ, and PIPKIγ, and rat PIPKIIα, PIPKIIβ, and PIPKIIγ were cloned into pEF-Bos-myc mammalian expression vectors as described previously [36]. The nomenclature of PIPKIα and PIPKIβ are different in mouse and human [27,28] and we use the human nomenclature in this study according to the revised nomenclature in the current GenBank database. Site-directed mutants of PIPKIα and PIPKIIα were generated using QuikChange II Site-Directed Mutagenesis Kit (Stratagene). Deletion mutants of PIPKIα and PIPKIIα were generated by subcloning truncated cDNA fragments of PIPKIα and PIPKIIα into pEF-Bos-myc. Bacterial expression vector encoding PIPKIα and PIPKIIα fused to glutathione S-transferase (GST) was generated by subcloning cDNA encoding PIPKIα and PIPKIIα into pGEX-4T-1 (GE healthcare). Green fluorescent protein (GFP)-tagged NBCe1-B was described previously [9].

Antibodies
Rabbit anti-IRBIT antibody was described previously [2]. Mouse anti-myc (9E10), goat anti- 15), and control goat IgG were obtained from Santa Cruz Biotechnology. Rat anti-HA (3F10) and mouse anti-myc antibody conjugated with peroxidase were obtained from Roche Applied Science. Donkey anti-rabbit IgG conjugated with horseradish peroxidase (HRP) and goat anti-rat IgG conjugated with HRP were obtained from GE healthcare. Rabbit anti-goat IgG conjugated with HRP was obtained from Medical & Biological Laboratories. Rabbit anti-HA, Alexa Fluor 594-conjugated goat anti-rabbit IgG, and Alexa Fluor 488 or Alexa Fluor 647-conjugated goat anti-mouse IgG were obtained from Invitrogen.

Immunoprecipitation
Fifteen to eighteen hours after transfection, COS-7 cells were lysed in lysis buffer (10 mM Hepes [pH 7.4], 100 mM NaCl, 2 mM EDTA, and 1% Nonidet P-40) for 30 min at 4°C, followed by centrifugation (20,000 x g, 30 min). The supernatants were incubated with protein G-Sepharose beads (GE healthcare) for 1 hr at 4°C to clarify nonspecific binding to the beads. The clarified supernatants were incubated with 0.5 μg anti-myc or 0.2 μg anti-HA antibody for 1 hr at 4°C, and immune complex was isolated by the addition of 5 μl protein G-Sepharose beads for 1hr at 4°C. The complex was spun down and washed three times with lysis buffer. Precipitated proteins were eluted by boiling in SDS-PAGE sample buffer and analyzed by immunoblotting with appropriate antibodies.
For immunoprecipitation from cerebellum, adult mouse cerebella were homogenized in 9 volumes of 10 mM Hepes buffer (pH 7.4) containing 320 mM sucrose, 2 mM EDTA, 1 mM 2-mercaptoethanol, and protease inhibitor cocktail (Roche Applied Science) with a glass-Teflon homogenizer (950 rpm, 10 strokes). The homogenate was centrifuged at 1,000 × g for 10 min, and the supernatant was centrifuged at 100,000 × g for 30 min to obtain cytosolic fraction (the supernatant) and membrane fraction (the pellet). The membrane fraction was solubilized in 50 mM Hepes, 2 mM EDTA, 1% Nonidet P-40, and protease inhibitors for 30 min at 4°C, followed by centrifugation (20,000 x g, 30 min). The supernatant was added with 100 mM NaCl and diluted two times with immunoprecipitation buffer (10 mM Hepes [pH 7.4], 100 mM NaCl, 2 mM EDTA, and 0.5% Nonidet P-40). The cytosolic and the solubilized membrane fractions were pre-cleared with protein G-Sepharose beads for 1 hr at 4°C, and the clarified supernatants were incubated with 2 μg anti-PIPK or control antibody pre-bound to 5 μl protein G-Sepharose beads for 2 hr at 4°C. The complex was spun down and washed three times with immunoprecipitation buffer. All animal studies were carried out in accordance with the guidelines and approval from the Animal Experiments Committee at the RIKEN Wako Institute (Approval number: H27-2-202).

Immunoblotting
Samples were separated by SDS-PAGE and transferred onto polyvinylidene fluoride membrane (Millipore) by electroblotting. After blocking with 5% skim milk in PBS containing 0.1% Tween 20 (PBS-T) for 1 h at room temperature, membranes were immunoblotted with primary antibodies diluted in PBS-T containing 3% skim milk at the appropriate concentrations (anti-HA, 0.1 μg/ml; anti-myc conjugated with peroxidase, 0.5 μg/ml; other antibodies, 1 μg/ ml) for 1 h at room temperature. After washing three times with PBS-T, the membrane was incubated with appropriate secondary antibodies conjugated with HRP diluted in PBS-T for 1 h at room temperature. When signals derive from goat antibodies used in the immunoprecipitation overlap those of immunoprecipitated proteins, the secondary antibody of the Immuno-Cruz TM IP/WB Optima D system (Santa Cruz Biotechnology) was used. After washing three times with PBS-T, immunoreactive bands were visualized with the ECL detection system (GE healthcare), and were captured using a luminescent image analyzer (LAS-4000 mini, GE healthcare). Band intensities were quantified using ImageJ software.

Measurement of PIPK activity
In vitro lipid kinase assays were performed as described previously [32,36] with modifications. Lysates of transfected COS-7 cells or MEF cells were incubated with 2 μg appropriate antibodies for 2 hr at 4°C. The immune complex was isolated by the addition of 10 μl protein G-Sepharose beads for 1hr at 4°C. The complex was spun down and washed three times with lysis buffer and three times with reaction buffer (50 mM Tris-HCl [pH 7.5], 100 mM NaCl, 1mM MgCl 2 , and 1mM EGTA). For PIPKIα activity, the reaction was started by adding 50 μM PI4P, 50 μM ATP, and 10 μCi of [γ-32 P]ATP (PerkinElmer Life Sciences) in 50 μl. For PIPKIIα activity, 50 μM PI5P was used instead of PI4P. After incubation for 10-20 min at 37°C, the lipids were extracted with chloroform/methanol and spotted onto TLC plates (Merck). The plates were developed in chloroform/methanol/ammonia/water, and the products were observed by autoradiography. PI(4,5)P 2 production was quantified by a liquid scintillation counter (Packard/PerkinElmer).

Immunocytochemistry
Transfected HeLa cells grown on glass coverslips were washed once with PBS, fixed in 4% formaldehyde in PBS for 15 min, permeabilized in 0.1% Triton X-100 in PBS for 5 min, and blocked in PBS containing 2% normal goat serum (Vector Laboratories) for 60 min at room temperature. Cells were then stained with rabbit anti-HA (1 μg/ml) and mouse anti-myc (1 μg/ ml) for 60 min at room temperature. Following four 5-min PBS washes, Alexa Fluor 594-conjugated goat anti-rabbit IgG (1 μg/ml) and Alexa Fluor 488 or 647-conjugated goat anti-mouse IgG (1 μg/ml) were applied for 45 min at room temperature. Following four 5-min PBS washes, the coverslips were mounted with Vectashield (Vector Laboratories) and observed under a confocal fluorescence microscopy (FV1000, Olympus) with a ×60 objective. Fluorescence images were analyzed by FV10-ASW software (Olympus). The average fluorescence intensity of the two cell edges were defined as the plasma membrane and regions between the cell edges were defined as cytosol.

IRBIT interacts with PIPKIα and PIPKIIα in cerebellum
We previously reported proteomic analysis of IRBIT-interacting proteins, where IRBIT was overexpressed in 293EBNA cells and proteins immunoprecipitated with IRBIT were analyzed by mass spectrometry [20]. Among the proteins identified was a protein belonging to the PIPK family, PIPKIIγ [32]. Here we investigated the interaction of IRBIT with the PIPK isoforms PIPKIα, PIPKIβ, PIPKIγ, PIPKIIα, PIPKIIβ, and PIPKIIγ (human nomenclature is used in this study). IRBIT and each isoform of the PIPK family were transfected into COS-7 cells, and coimmunoprecipitation assays were performed. IRBIT interacted with all PIPK isoforms in the heterologous expression system (Fig 1A, S1 Fig).
Further, we tested the interaction of IRBIT with PIPK family members in vivo. Mouse cerebella, where IRBIT is highly expressed [2,40], were fractionated into cytosolic and membrane fractions and processed for immunoprecipitation using antibodies specific to each PIPK isoform. IRBIT was clearly detected in the anti-PIPKIα immunoprecipitates from the cytosol fraction and in the anti-PIPKIα and anti-PIPKIIα immunoprecipitates from the membrane fraction ( Fig 1B, S1 Fig). Faint evidence of IRBIT was observed in the anti-PIPKIγ immunoprecipitates from the membrane fraction ( Fig 1B, S1 Fig). These results indicate that isoform specificity exists in vivo, and IRBIT mainly binds to PIPKIα and PIPKIIα in mouse cerebellum. The restriction of IRBIT/PIPKIIα interaction to the membrane fraction might be accounted for by the observation that IRBIT is more highly phosphorylated in membrane than in cytosolic fractions [40]. Although these results do not exclude the possibility that IRBIT interacts with other PIPK isoforms in different tissues, we hereafter focused on its interaction with PIPKIα and PIPKIIα.
Conserved catalytic amino acids of PIPKIα and PIPKIIα are required for interaction with IRBIT PIPK family members have a conserved kinase homology domain and divergent N-and C-terminal regions [25,27] (Fig 2A). To identify the regions responsible for binding to IRBIT, we analyzed the interaction of IRBIT with deletion mutants of PIPKIα and PIPKIIα. The C-terminal deletion mutants PIPKIα-(1-507) and PIPKIα-(1-438) bound to IRBIT, whereas PIPKIα-(1-430) showed significantly decreased affinity to IRBIT (Fig 2A and 2B). The expression levels of the N-terminal truncation mutants, PIPKIα-(57-546) and PIPKIα-(97-546), were much lower than that of wild-type PIPKIα (Fig 2A and 2B). These results suggest that amino acids 431-438 of PIPKIα are critical for the interaction with IRBIT, and the N-terminal divergent regions are essential for the stability of PIPKIα. Deletion of the N-terminal region of PIPKIIα resulted in low expression and reduced interaction with IRBIT (Fig 2A and 2C).
Although the homology of primary sequences of the kinase domains of type I and type II PIPKs is not very high (30%-40% identity) [27][28][29][30][31][32], their 3D structures are assumed to be well conserved [41,42]. Considering that all PIPK isoforms bound to IRBIT in a heterologous expression system (Fig 1A, S1 Fig), the conserved kinase homology domain is expected to be involved in the interaction with IRBIT. Our previous study showed that IRBIT binds to the IP 3 -binding core of IP 3 R via amino acids essential for IP 3 recognition [2,3]. Similarity in chemical structures between IP 3 and the PIPK substrates, PI4P and PI5P, both having an inositol ring and two or three phosphoryl groups, prompted us to investigate whether IRBIT interacts with the catalytic core of PIPKIα and PIPKIIα. Structural homologies between PIPK family members and protein kinases and mutagenesis analysis showed that Lys179, Asp307, and Asp389 of PIPKIα, and the corresponding Lys145, Asp273, and Asp359 of PIPKIIα are located in the catalytic core and are essential for enzymatic activity [29,[41][42][43] (Fig 2A). These residues are completely conserved among PIPK family members (Fig 2D). We introduced mutations in Lys179, Asp307, and Asp389 of PIPKIα and tested their interaction with IRBIT. Co-immunoprecipitation assays revealed that the D389A mutant displayed slightly decreased affinity for IRBIT, and that the double mutation D307A/D389A further decreased the interaction with IRBIT (Fig 2E). To exclude the indirect effect of actin re-organization induced by PIPKIα overexpression [44] on this interaction, we performed pull-down assays using recombinant GST fusion proteins of PIPKIα mutants. We found that the D307A and D389A mutants exhibited reduced binding to IRBIT, and the double mutant D307A/D389A displayed further diminished interactions (Fig 2F). Similar results were obtained for PIPKIIα. Mutation of Asp273 or Asp359 of PIPKIIα, corresponding to Asp307 and Asp389 of PIPKIα, respectively, slightly decreased PIPKIIα binding to IRBIT, and the double mutation D273A/D359A markedly weakened this interaction (Fig 2G). Thus, aspartate residues, Asp307/Asp389 and Asp273/Asp359, which are critical for the kinase activity of PIPKIα and PIPKIIα, respectively, are essential for To clarify further the possibility that IRBIT binds to the catalytic core of PIPKs, we tested whether the PIPK substrates, ATP, and Mg 2+ , which interact with the catalytic core of PIPKs, interfere with the interaction between IRBIT and PIPKIα/IIα. Pull-down assay using recombinant IRBIT showed that the binding of IRBIT to PIPKIα was suppressed by PI4P and 5 mM ATP, but not by PI5P, IP 3 , ADP, and TTP (Fig 3A). On the other hand, the interaction between IRBIT and PIPKIIα was efficiently inhibited by 1-5 mM ATP, but not by PI4P, PI5P, IP 3 , ADP, and TTP, in a dose-dependent manner (Fig 3B). As a control, IP 3 disrupted the interaction between IRBIT and the IP 3 binding domain (IP 3 BD) of IP 3 R (Fig 3C), as reported previously [2,3]. A high concentration of Mg 2+ (10 mM) reduced the binding of IRBIT to PIPKIα, but not to PIPKIIα, although physiological concentrations of Mg 2+ (0.25-1 mM) [45] did not affect the interactions (Fig 3D). These results indicate that substances that bind to the catalytic core of PIPKs interfere with the IRBIT binding, supporting the hypothesis that IRBIT interacts with the catalytic core of PIPKIα and PIPKIIα. The mechanism of the interactions is different between PIPKIα and PIPKIIα, that is, the PIPKIα-IRBIT interaction is sensitive to PI4P, ATP, and Mg 2+ , whereas the PIPKIIα-IRBIT interaction is resistant to PI4P/PI5P and Mg 2+ , and is more sensitive to ATP than the PIPKIα-IRBIT interaction. Different affinities of the interactions and/or involvements of non-conserved amino acids in the catalytic core of PIPKIα and PIPKIIα might explain this difference. Intracellular ATP levels are approximately 1 mM in mammalian cells [46][47][48] and can be increased to several millimoles by extracellular stimuli [46,49]. Thus, the interaction between IRBIT and PIPKs might be regulated by changes of cytosolic ATP concentrations.
Several lines of evidence suggest that IRBIT behaves as if its moiety mimics the structure or function of inositol phosphates or phosphoinositides. First, IRBIT binds to the IP 3 -binding core of IP 3 R by recognizing amino acids that interact with IP 3 [3,50]. Second, IRBIT and PI (4,5)P 2 bind to the same domain of NBCe1-B and activate NBCe1-B, in part by a common mechanism [13]. Third, IRBIT binds to PIPKIα and PIPKIIα via amino acids that catalyze phosphatidylinositol phosphates (Fig 2D-2G), and PI4P inhibits the IRBIT-PIPKIα interaction (Fig 3A). These findings suggest that phosphorylated serine/threonine residues of IRBIT (discussed below) simulate phosphate groups of inositol phosphates and phosphoinositides.
cerebellum. Cytosolic and membrane fractions were processed for immunoprecipitation with antibodies specific to each PIPK isoform or control (Ctrl) antibody, and immunoprecipitates were analyzed by Western blotting with antibodies indicated. Anti-PIPKIIγ antibody that successfully precipitates PIPKIIγ was not available. Asterisks indicate immunoglobulin heavy chains. doi:10.1371/journal.pone.0141569.g001 Several serine/threonine residues in the serine-rich region of IRBIT are phosphorylated. Ser68 is phosphorylated by protein kinase D [3,7,17]. Once Ser68 is phosphorylated, Ser71,    E). HA-IRBIT mutants were transfected with myc-PIPKIα or myc-PIPKIIα into COS-7 cells and were processed for immunoprecipitation with anti-HA or anti-myc antibody. Immunoprecipitates were analyzed by Western blotting with anti-HA or anti-myc antibody. (F) Quantification of (D) and (E). The relative binding of site-directed mutants of IRBIT to PIPKIα and PIPKIIα are presented as mean +/-SEM (n = 4). * p < 0.05, ** p < 0.01, compared with wild-type (WT) IRBIT (one-way ANOVA with post-hoc Tukey's HSD test). (G) Prediction of intrinsically disordered regions of IRBIT. Amino acid sequence of IRBIT was analyzed by three different meta-protein disorder prediction programs, metaDisorder [37], PONDR-FIT [38], and metaPrDOS [39]. Disordered regions predicted by three (red), two (orange), or one (yellow) programs are indicated. Ser74, and probably Ser77 are subsequently phosphorylated by casein kinase I [3,7]. Thr82, Ser84, and Ser85 are phosphorylated in mouse brain [8]. Thus, we analyzed the interaction of site-directed mutants of serine/threonine residues in the IRBIT serine-rich region with PIPKIα and PIPKIIα. Mutation of Ser68 or Ser71 had the most prominent effect, showing almost complete loss of interaction with both PIPKIα and PIPKIIα (Fig 4D-4F). Mutation of Ser74, Thr82, or Ser84 markedly reduced the binding to PIPKIα, but had little effect on the binding to PIPKIIα. Mutation of Ser66, Ser70, or Ser80 also significantly decreased the interaction with PIPKIα, but not with PIPKIIα. In contrast, mutation of Ser77 profoundly decreased binding to PIPKIIα, but had less effect on the interaction with PIPKIα (Fig 4D-4F). Thus, Ser68 and Ser71 are commonly required for the interaction with PIPKIα and PIPKIIα, and Ser66/Ser70/ Ser74/Ser80/Thr82/Ser84 and Ser77 are specifically important for the binding to PIPKIα and PIPKIIα, respectively. These results suggest that the phosphorylation status of the serine-rich region determines the binding specificity of IRBIT to PIPK isoforms. Furthermore, differences in the phosphorylation status of IRBIT between the cytosol and membrane fractions [2,40] may explain the result that IRBIT bound to PIPKIIα in the membrane fraction but not in the cytosolic fraction (Fig 1B, S1 Fig).

The N-terminal region of IRBIT is intrinsically highly disordered
The serine-rich region of IRBIT is critically involved in its interaction with most of its binding partners [2,3,9,10,16,17,[20][21][22]. The molecular mechanisms by which the serine-rich region interacts with diverse proteins that have no sequence or structural homology remain largely unknown. Circular dichroism spectroscopy and limited proteolysis analyses suggest that the N-terminal domain of IRBIT has a loose structure [51]. In addition, the crystal structure of IRBIT available in the Protein Data Bank (PDB number 3MTG) lacks the N-terminal 100 amino acids [52]. These observations suggest that the regions surrounding the serine-rich region have flexible structures. Thus, we investigated the structural order/disorder of IRBIT using meta-protein disorder prediction programs [37][38][39]. The in silico predictions revealed that the N-terminal approximately 100 amino acids of IRBIT containing the serine-rich region is highly disordered (Fig 4G), indicating that the N terminus of IRBIT belongs to the intrinsically disordered protein (IDP) regions [53].
The IDP regions do not form stable structures, and their instability is encoded by amino acid sequences [53]. The conformational flexibility of the IDP regions is beneficial for interactions with diverse proteins [54]. The structure and function of the IDP regions is regulated by phosphorylation [55,56]. The structural flexibility of the serine-rich region of IRBIT may underlie the mechanism of its interactions with multiple divergent proteins and its regulation by phosphorylation, and may allow for its intrusions into the catalytic core of PIPKs (Fig 2D-2G) and the IP 3 binding core of IP 3 R [3]. The IRBIT gene is assumed to emerge by gene duplication of an ancestral AHCY gene in the course of evolution [6]. The key roles of the serinerich region in IRBIT function suggest that acquisition of the N-terminal IDP region confers IRBIT with crucial abilities to regulate diverse cellular processes required in higher organisms.

IRBIT does not regulate the kinase activity of PIPKIα and PIPKIIα
We tested whether IRBIT regulates the enzymatic activity of PIPKIα and PIPKIIα. In vitro production of PI(4,5)P 2 was measured using immunoprecipitates from COS-7 cells transfected with PIPKIα with or without IRBIT. IRBIT had no effect on the kinase activity of PIPKIα ( Fig  5A). Similar results were obtained for PIPKIIα (Fig 5B). Further, we analyzed the activity of PIPKIα and PIPKIIα in MEF cells derived from IRBIT knockout mice [22]. Expression of PIP-KIα and PIPKIIα did not differ between wild-type and IRBIT knockout MEF cells (Fig 5C). In vitro lipid kinase assays of immunoprecipitates with anti-PIPKIα or anti-PIPKIIα antibodies showed that IRBIT knockout did not influence PI(4,5)P 2 production (Fig 5D and 5E). These results suggest that IRBIT does not regulate the lipid kinase activity of PIPKIα and PIPKIIα.
The interaction of IRBIT with the catalytic core of PIPKIα and PIPKIIα (Fig 2D-2G, Fig 3) may account for the inability of IRBIT to regulate the enzymatic activity of PIPKs. That is, the PI4P and/or ATP used in the in vitro lipid kinase assays might competitively disrupt the interaction between IRBIT and PIPKs. Therefore, we cannot exclude the possibility that IRBIT might regulate PIPK activity when the substrate concentration is extremely low.

IRBIT and PIPKs interact with NBCe1-B
We tested whether IRBIT and PIPKs form complexes with PI(4,5)P 2 effector proteins. IRBIT interacts with and activates the Na + /HCO 3 − cotransporter NBCe1-B [9][10][11][12][13][14], an ion transporter that mediates Na + -dependent HCO 3 − transport across the plasma membrane [57]. NBCe1-B activity is regulated by PI(4,5)P 2 [13,35]. Thus, we examined the binding of PIPKIα to NBCe1-B by pull-down assay using the N-terminal cytoplasmic domain of NBCe1-B (NBCe1-B/N) [9]. Further, we tested the interaction of PIPKIα with IP 3 R using the IP 3 binding domain (IP 3 BD) of IP 3 R [2], as IP 3 , the agonist of IP 3 R, is generated by the hydrolysis of PI (4,5)P 2 . As shown in Fig 6A, PIPKIα interacted with NBCe1-B/N but not with IP 3 BD, although IRBIT bound to both NBCe1-B/N and IP 3 BD, as reported previously [2,9]. These results suggest that IRBIT, PIPKIα, and NBCe1-B form a ternary complex. Further, we examined the interaction of PIPKIIα with NBCe1-B or IP 3 R. IRBIT enhanced the interaction of PIPKIIα with NBCe1-B/N but not with IP 3 BD (Fig 6B), suggesting that IRBIT recruits PIPKIIα to NBCe1-B. Next, we investigated intracellular localization of IRBIT, PIPKIα, PIPKIIα, and NBCe1-B in HeLa cells. When overexpressed alone, IRBIT and PIPKIIα were localized in the cytosol, whereas PIPKIα was localized at the plasma membrane (Fig 7A). Cells overexpressing PIPKIα showed a rounded cell shape due to actin re-organization induced by PI(4,5)P 2 production [29,44] (Fig 7A). NBCe1-B was localized at the plasma membrane and intracellular vesicles ( Fig  7A). When IRBIT was co-expressed with PIPKIα, or NBCe1-B, but not with PIPKIIα, IRBIT was partially localized near the plasma membrane (Fig 7B and 7C). IRBIT co-localized with NBCe1-B at the plasma membrane, but not on intracellular vesicles (Fig 7Bc). The effects of PIPKIα and NBCe1-B on the plasma membrane targeting of IRBIT might be additive (Fig 7B  and 7C). IRBIT, PIPKIα, and NBCe1-B colocalized at the plasma membrane (Fig 7Bd). These results suggest that IRBIT, PIPKIα, and NBCe1-B form signaling complexes at the plasma membrane.
The physiological role of the complexes containing IRBIT, PIPKIα and NBCe1-B is unclear. IRBIT directly activates NBCe1-B [9][10][11][12][13][14], whereas PIPKIα may indirectly potentiates the activity of NBCe1-B through the production of PI(4,5)P 2 in close proximity to NBCe1-B. IRBIT is most highly expressed in the brain [2,40], and IRBIT knockout mice show behavioral abnormities such as hyperactivity and increased social interactions [22]. PIPKIα is abundantly expressed in the brain and is responsible for the generation of specific PI4,5,P 2 pools [58]. In the brain, NBCe1-B plays a crucial role in the regulation of intracellular pH in neurons and astrocytes [59][60][61][62]. In fact, human patients with mutations in the NBCe1 gene (SLC4A4) show neurological abnormalities such as mental retardation and migraines [60,[63][64][65][66]. Thus, the signaling complex containing IRBIT, PIPKIα, and NBCe1-B may play an important role in the central nervous system. Compared to type I PIPKs, type II PIPKs play a minor role in PI(4,5) P 2 production, as the basal level of PI5P is only 0.5%-2% that of PI4P [67]. Therefore, the major function of type II PIPKs may be to control PI5P levels [68]. PI5P is the least characterized of the phosphoinositides; however, one of its functions is speculated to be the regulation of intracellular vesicle trafficking [68][69][70]. Thus, the IRBIT-PIPKIIα complex may regulate NBCe1-B trafficking. In the cerebellum, IRBIT bound to PIPKIIα in the membrane fraction, but not in the cytosol (Fig 1B, S1 Fig); however, IRBIT colocalized with cytosolic PIPKIIα in HeLa cells (Fig 7Bb). This discrepancy may be explained by the abundant expression of NBCe1-B and its splicing variant, NBCe1-C, in brain [59,60], and/or difference in phosphorylation status of IRBIT between brain and HeLa cells. Further studies are required to reveal physiological roles of the interaction between IRBIT, PIPKIα/IIα, and NBCe1-B.

Conclusions
In this study, we demonstrated that IRBIT interacts with PIPKs, lipid kinases that generate PI (4,5)P 2 . Specifically, IRBIT bound to PIPKIα and PIPKIIα isoforms in the cerebellum. Mutagenesis analysis and inhibition of the interaction by PI4P, Mg 2+ and/or ATP suggest that IRBIT interacts with the catalytic core of PIPKIα and PIPKIIα through conserved catalytic aspartate residues. Phosphorylation sites in the serine-rich region of IRBIT are essential for the selectivity of its interaction with PIPKIα and PIPKIIα, suggesting the possibility that phosphorylated serine/threonine residues of IRBIT might simulate phosphate groups of phosphatidylinositol phosphate substrates. The structural flexibility of the N-terminal region of IRBIT may underlie the mechanism of its interactions with multiple divergent proteins. Furthermore, in vitro binding experiments and immunocytochemistry suggest that IRBIT and PIPKIα interact with the Na + /HCO 3 − cotransporter NBCe1-B. These results suggest that IRBIT forms signaling complexes with PIPKIα and NBCe1-B, whose activity is regulated by PI(4,5)P 2 .
Supporting Information S1 Fig. IRBIT interacts with PIPKs in the presence of 1 mM Mg 2+ . (A) Immunoprecipitation from heterologous cells using the lysis buffer containing 1 mM MgCl 2 and 1 mM EGTA instead of 2 mM EDTA. HA-IRBIT and myc-PIPK isoforms transfected into COS-7 cells were immunoprecipitated with anti-HA or anti-myc antibody. Immunoprecipitates were analyzed by Western blotting with anti-HA or anti-myc antibody. (B) Immunoprecipitation from mouse cerebellum using buffers containing 1 mM MgCl 2 and 1 mM EGTA instead of 2 mM EDTA. Cytosolic and membrane fractions were processed for immunoprecipitation with antibodies specific to each PIPK isoform or control (Ctrl) antibody, and immunoprecipitates were analyzed by Western blotting with antibodies indicated. Asterisks indicate immunoglobulin heavy chains. (TIF)