Inositol hexakisphosphate kinase 1 (IP6K1) activity is required for cytoplasmic dynein-driven transport

Inositol pyrophosphates, such as diphosphoinositol pentakisphosphate (IP7), are conserved eukaryotic signaling molecules that possess pyrophosphate and monophosphate moieties. Generated predominantly by inositol hexakisphosphate kinases (IP6Ks), inositol pyrophosphates can modulate protein function by posttranslational serine pyrophosphorylation. Here, we report inositol pyrophosphates as novel regulators of cytoplasmic dynein-driven vesicle transport. Mammalian cells lacking IP6K1 display defects in dynein-dependent trafficking pathways, including endosomal sorting, vesicle movement, and Golgi maintenance. Expression of catalytically active but not inactive IP6K1 reverses these defects, suggesting a role for inositol pyrophosphates in these processes. Endosomes derived from slime mold lacking inositol pyrophosphates also display reduced dynein-directed microtubule transport. We demonstrate that Ser51 in the dynein intermediate chain (IC) is a target for pyrophosphorylation by IP7, and this modification promotes the interaction of the IC N-terminus with the p150Glued subunit of dynactin. IC–p150Glued interaction is decreased, and IC recruitment to membranes is reduced in cells lacking IP6K1. Our study provides the first evidence for the involvement of IP6Ks in dynein function and proposes that inositol pyrophosphate-mediated pyrophosphorylation may act as a regulatory signal to enhance dynein-driven transport.


Introduction
Inositol pyrophosphates are small-molecule second messengers composed of an inositol ring containing pyrophosphate groups in addition to monophosphates [1,2]. They occur ubiquitously in all eukaryotes and participate in many cellular processes, including DNA repair, stress response, apoptosis, phosphate metabolism, and energy homeostasis [3]. 5-Diphosphoinositol pentakisphosphate (5-IP 7 ) is the most abundant inositol pyrophosphate in mammalian cells with cellular concentrations ranging from 0.5 to 3 mM [4]. 5-IP 7 is synthesized from inositol hexakisphosphate (IP 6 ) by inositol hexakisphosphate kinases (IP6Ks), which have three homologs in mammals, such as IP6K1, IP6K2, and IP6K3 [5][6][7]. Other inositol pyrophosphates, such as 1-diphosphoinositol pentakisphosphate (1-IP 7 ) and 1,5-bis-diphosphoinositol tetrakisphosphate (IP 8 ), are generated by another class of enzymes called PP-IP5 kinases [7] and occur in lower amounts in most eukaryotic cells [4]. Inositol pyrophosphates can modulate protein function in two ways: (a) by direct binding to a target protein or (b) by conferring a posttranslational modification known as pyrophosphorylation [8,9]. The latter mechanism involves the transfer of a high-energy β-phosphate from an inositol pyrophosphate such as 5-IP 7 to a phosphorylated serine residue to form pyrophosphoserine and has been shown to regulate protein-protein interactions [10,11].
In addition to their extensive role in regulating metabolic and signaling pathways, inositol pyrophosphates have been shown to participate in different vesicle trafficking processes. In yeast, vesicle transport takes place along the actin cytoskeleton with the help of myosin motors [12]. In mammals, short-range movement of endocytic and exocytotic vesicles near the plasma membrane is also actin-myosin 10 ml of Dulbecco's phosphate-buffered saline (PBS, Life Technologies). Macrophages were maintained for 24-48 h in the same medium as MEFs prior to experiments.

Immunofluorescence
Cells grown on coverslips were fixed with 4% paraformaldehyde (PFA), permeabilized in 0.1% Triton-X 100 for 5 min, and incubated in blocking solution (2% BSA in PBS) for 1 h at room temperature. Cells were then incubated for 2-18 h in primary antibodies diluted appropriately in blocking solution, followed by incubation with secondary antibodies diluted in blocking solution for 1 h. Coverslips were mounted on glass slides using mounting medium containing DAPI (H-1200, Vector Labs). Images were acquired using an LSM 510 (LSM acquisition software) or LSM 700 (Zen acquisition software) confocal microscope (Zeiss) equipped with 405, 488, and 555/561 nm lasers and fitted with a ×63, 1.4 N.A. objective.

Fluorescent ligand uptake and trafficking assays
Tfn endocytosis and trafficking assays were done as previously described [21] with slight modifications. To monitor Tfn endocytosis by flow cytometry, MEFs grown in 35 mm dishes were serum-starved for 30 min in 0.5% BSA-containing DMEM, followed by 5 min incubation with 25 mg/ml Alexa488 Tfn at 37°C. Cells were washed with cold PBS, trypsinized, and transferred to chilled tubes containing DMEM. The cells were pelleted by centrifugation and resuspended in 3% PFA. At least 10 000 cells were analyzed by flow cytometry (BD Accuri C6) using a 488 nm laser. For microscopy, MEFs grown on glass coverslips were incubated in serum-free medium for 1 h at 37°C, followed by 25 mg/ml Alexa488 Tfn on ice for 30 min. Cells were allowed to take up the bound Alexa488 Tfn for 5 min at 37°C to monitor endocytosis or for 1 h at 37°C to monitor accumulation in the endocytic recycling compartment (ERC). Cells were washed with chilled Dulbecco's PBS (Life Technologies), fixed using 4% PFA, and, where required, subjected to IF with EEA1 antibody. To measure CT-B binding to the plasma membrane, MEFs were serum-starved for 1 h at 37°C and then incubated with 5 mg/ml Alexa Fluor 594 CT-B for 1 h on ice. Cells were washed with Dulbecco's PBS and fixed with 4% PFA. Coverslips were mounted in mounting medium containing DAPI (H-1200, Vector Labs), and images were acquired as above.

Image analysis
Where indicated, images were subjected to adjustment of tonal range on the whole image using Adobe Photoshop (levels adjustment) to improve visualization for representation purposes. Such adjustments were identical for all images in a single assay. For quantification, raw images were analyzed. To assess Tfn uptake and CT-B binding, fluorescence intensity per cell was quantified using the ImageJ 'measure' tool. Individual cells were marked using the 'selection tool' and the integrated density value of the fluorescence signal in each cell was recorded. Cells with accumulation of Tfn in ERC-like structures were scored as described previously [21]. Analysis of Golgi morphology was performed as described earlier [22]. Colocalization analysis was performed in ZEN 2012 SP1 software (Zeiss). Briefly, a fixed threshold was set across all images to exclude background fluorescence and the percentage of colocalization of EEA1 with Tfn in the entire cell was calculated by the ratio of the number of colocalized pixels of EEA1 to the total EEA1 pixels.

Live cell imaging
MEFs grown on glass-bottom dishes were serum-starved for 1 h and incubated with 5 mg/ml Alexa Fluor 594 CT-B for 30 min at 37°C in serum-free medium. After adding complete medium, cells were placed in an incubation chamber with controlled temperature (37°C) and CO 2 (5%) on an inverted Zeiss LSM510 confocal microscope. Images were acquired at intervals of 10 s for 200 s using the 561 nm laser with a ×63 1.4 N.A.
objective. CT-B vesicles were tracked over several frames using the ImageJ 'manual tracking' plugin. Only vesicles that were distant from the nucleus and plasma membrane were tracked.

Phagosome distribution assay
Peritoneal macrophages derived from Ip6k1 +/+ and Ip6k1 −/− mice grown on glass-bottom dishes were serumstarved for 1 h and incubated with 750 nm carboxylated latex beads (07759, Polysciences) for 1 h to allow phagocytosis. The unphagocytosed beads were washed with serum-containing media, and the cells were incubated for 1 h in the same medium at 37°C. Differential interference contrast (DIC) images were acquired using a live cell multipoint imaging system (Nikon Ti Eclipse) equipped with a ×100 1.4 N.A. objective, ×1.5 intermediate magnification, and a CoolSnap HQ CCD camera. Images were analyzed by ImageJ to identify the cell and nuclear boundaries and to examine the movement of phagocytosed beads along the long edge of the cell towards the nucleus. The fractional distance of each bead from the nuclear centroid was calculated as the ratio of the distance of the bead from the center of the nucleus to the distance of the cell membrane from the center of the nucleus along the line joining the centroid and bead.

Endosome motility analysis in D. discoideum
Endosome motility in D. discoideum was performed essentially as described earlier [23]. Briefly, 4-6 × 10 8 cells of wild-type and i6kA − amoebae were harvested, washed with Sorenson's phosphate buffer ( pH 6.0), resuspended in 1:1 (w/v) lysis buffer [30 mM Tris-HCl, pH 8.0, 4 mM EGTA, 3 mM DTT, 5 mM benzamidine HCl, 5 mM PMSF containing 30% (w/v) sucrose and protease inhibitor cocktail (Roche)], and lysed by passing through a polycarbonate filter with 5 mm pore size. The crude lysate was centrifuged at 2000 g for 5 min at 4°C to obtain the postnuclear supernatant (PNS). PNS (0.5 ml) was diluted with 18.5 ml of lysis buffer and 1 ml of ATP regeneration system (1 mM ATP, 1 mM MgCl 2 , 2 mM creatine phosphate, and 2 U/ml creatine kinase). This motility mixture was passed into a flow cell containing polarity-labeled microtubules. Optical trap experiments were performed to assess the direction of motility of refractile endosomes by customized video-enhanced DIC microscopy (Nikon) using a ×100 oil immersion objective. Motion of single endosomes was recorded at 30 frames/s and tracked offline with ∼5 nm resolution, followed by analysis of motion using Bayesian optimization to extract velocities of motion [24].

Protein purification, phosphorylation, and pyrophosphorylation
GST-tagged dynein IC fragments corresponding to residues 1-70 and 1-111 were expressed in Escherichia coli BL21(DE3) strain and purified using glutathione-agarose beads (GE Life Sciences) by standard procedures. Radiolabeled IP 7 synthesis, CK2-mediated phosphorylation, and IP 7 -mediated pyrophosphorylation were performed as described earlier [9]. For pyrophosphorylation assays, beads were first treated with CK2 enzyme (New England Biolabs) in protein kinase buffer (New England Biolabs) and 0.5 mM Mg 2+ -ATP for 30 min at 30°C, then washed with cold PBS, resuspended in IP 7 pyrophosphorylation buffer (25 mM HEPES, pH 7.4, 50 mM NaCl, 6 mM MgCl 2 , and 1 mM DTT) containing 5-7 mCi 5[β-32 P]IP 7 , and incubated at 37°C for 15 min. LDS sample buffer (NP0008, Life Technologies) was added to the beads, and the sample was heated at 95°C for 5 min. Proteins were resolved on a 4-12% NuPAGE Bis-Tris gel (Thermo Fisher Scientific), transferred to a PVDF membrane (GE Life Sciences), and pyrophosphorylation was detected using a phosphorimager (Typhoon FLA-9500). The amount of protein loaded was quantified based on the intensity of Ponceau S staining, which has been shown to be linear with increasing protein amount up to 140 mg [25]. Radiolabeled protein as a fraction of total protein was quantified using ImageJ.
For back-phosphorylation assays, dynein IC was immunoprecipitated from Ip6k1 +/+ and Ip6k1 −/− MEFs. Protein on beads was subjected to CK2-mediated phosphorylation by incubating with CK2 enzyme (New England Biolabs) in protein kinase buffer (New England Biolabs) in the presence of 0.5 mM Mg 2+ -ATP and 1-2.5 mCi [γ 32 -P]ATP for 30 min at 30°C. For back-pyrophosphorylation assays, beads were incubated with 5 [β-32 P]IP 7 as above, but without CK2 pre-phosphorylation. Radiolabeled proteins were detected using a phosphorimager (Typhoon FLA-9500), and total protein was subsequently detected by IB. Care was taken to ensure that the chemiluminescence signal was below saturation level. Radiolabeled protein as a fraction of total immunoprecipitated protein was quantified using ImageJ. We observed a variation in the apparent molecular weight of dynein IC in NuPAGE Bis-Tris gels compared with SDS-PAGE Tris-Glycine gels.

Mass spectrometry identification of phosphosites
GST-tagged IC-2C fragments were purified from E. coli, phosphorylated by CK2 as described above but without the incorporation of radiolabeled ATP, and resolved on a 4-12% Nu-PAGE Bis-Tris gel. Bands were visualized with Simply Blue Safe Stain (Invitrogen), excised, and sent to the Taplin Biological Mass Spectrometry Facility (TMSF), Harvard Medical School, Boston, USA for phosphosite identification. Briefly, the gel pieces were diced and subjected to a modified in-gel trypsin digestion and peptide extraction [26]. Samples were loaded via an FAMOS autosampler (LC Packings) onto a packed C18 reverse-phase HPLC capillary column (100 mm inner diameter × ∼30 cm length) and resolved with an acetonitrile gradient (2.5-97.5% acetonitrile and 0.1% formic acid). Each eluted peptide was subjected to electrospray ionization and entered into an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo Fisher Scientific). Detection, isolation, and fragmentation of eluted peptides were conducted to generate a tandem mass spectrum of fragment ions specific for each peptide. The acquired fragmentation pattern was matched to translated nucleotide or protein databases using SEQUEST (Thermo Finnigan) to determine the corresponding peptide sequence. The modification of 79.9663 mass units to Ser, Thr, and Tyr was included in the database searches to identify phosphopeptides. Phosphorylation assignments were determined by the Ascore algorithm [27]. According to instructions from TMSF, a phosphopeptide with an Ascore of >19 was considered to be phosphorylated with 99% certainty.

Protein interaction studies
To monitor interaction of the dynein IC N-terminus with endogenous p150 Glued , 5 mg of GST-tagged dynein IC(1-111) bound to glutathione-agarose beads was phosphorylated with CK2 and unlabeled Mg 2+ -ATP, and incubated with 50 mM 5-IP 7 , 5-PCP-IP 5 , or IP 6 at 37°C for 15 min, and then at 55°C for 20 min. Beads were washed with lysis buffer (25 mM HEPES, pH 7.4, 1% NP-40, 0.1% BSA, 100 mM KCl, 150 mM NaCl, and protease and phosphatase inhibitor cocktails) and incubated overnight with 1 mg/ml HEK293T lysate (prepared in lysis buffer). Bound proteins were eluted in Laemmli buffer and were resolved by 10% SDS-PAGE. Following western transfer to a PVDF membrane, the proteins were detected by IB with anti-GST and anti-p150 Glued antibodies. For coimmunoprecipitation of endogenous dynein IC and p150 Glued , MEFs harvested from 10 cm dishes were lysed by shearing in lysis buffer without NP-40 or by solubilization for 30 min at 4°C in NP-40 containing lysis buffer. The lysate was centrifuged at 14 000 g for 20 min. Proteins were cross-linked for 30 min by the addition of 2.5 mM DSP, a thiolcleavable cross-linker (22585, Thermo Scientific). The cross-linker was added either during cell lysis or subsequent to the removal of the cell debris, and cross-linking was quenched by adding 50 mM Tris-Cl, pH 7.4, for 15 min at 4°C. Two micrograms of anti-dynein IC or anti-p150 Glued antibodies were added to the lysate and incubated overnight at 4°C. Protein A/G Dynabeads (88802, Thermo Scientific) were added to the complex and incubated at 4°C for 2 h. Proteins were eluted in Laemmli buffer containing 179 mM β-mercaptoethanol to reverse DSP cross-links and analyzed by IB with anti-dynein IC or anti-p150 Glued antibodies.

Crude membrane preparation
Ip6k1 +/+ and Ip6k1 −/− MEFs grown in 150 mm dishes were scraped in 2 ml of fractionation buffer (25 mM HEPES, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 250 mM sucrose, 1 mM DTT, and protease inhibitor cocktail). Cells were lysed by passing through a 24-gauge needle (total homogenate, TH) and centrifuged at 10 000 g to pellet nuclei and mitochondria. The supernatant containing the cytosol and the membrane fraction (PNS) was subjected to high-speed centrifugation (Beckman) at 100 000 g for 1 h. The pellet containing the membrane fraction (membrane pellet, MP) was solubilized in 500 ml of fractionation buffer containing 1% Triton-X-100. The subcellular protein fractions were resolved on a 4-12% Nu-PAGE gel and analyzed by IB with anti-dynein IC, anti-p150 Glued , anti-GM130, and anti-α-tubulin antibodies.

Statistics
Graphs and analyses were produced using GraphPad Prism 5. Continuous data were tested for normality and analyzed using one-way ANOVA, two-tailed unpaired Student's t-test, or two-tailed Mann-Whitney test, as appropriate. Fisher's exact test was used to assess categorical data presented as contingency tables. Statistical significance was defined as P < 0.05. The cell numbers used to obtain quantitative data (n) and the number of independent experiments performed are indicated in the figure legends.

Endosomal sorting of Tfn in fibroblasts requires IP6K1 activity
To investigate the role of inositol pyrophosphates in dynein-dependent vesicle transport, we used MEFs derived from Ip6k1 −/− mice, in which IP 7 levels are reduced to ∼30% compared with Ip6k1 +/+ MEFs [20]. In interphase cells, dynein mediates endosomal sorting and Golgi organization [28]. To examine endosomal sorting, we tracked the localization of the iron-binding protein Tfn, which undergoes clathrin-dependent endocytosis and is transported from early endosomes to recycling endosomes in a dynein-dependent manner [28]. We allowed cells to take up fluorescently tagged Tfn for 1 h and scored for cells with accumulation of Tfn in the perinuclear ERC. The fraction of cells containing Tfn in ERC-like structures was significantly reduced in Ip6k1 −/− compared with Ip6k1 +/+ MEFs ( Figure 1A,B). When catalytically active IP6K1 was expressed in Ip6k1 −/− MEFs, it restored the subcellular distribution of Tfn, whereas expression of similar levels of the kinase-dead IP6K1 mutant enzyme (IP6K1 K226A/S334A ) [17] had no effect ( Figure 1A,B). This suggests that the synthesis of inositol pyrophosphates by IP6K1 is required for efficient Tfn trafficking in cells. To rule out any defect in early endocytosis, we pulse-labeled cells with Tfn for 5 min and quantified the endocytosed Tfn by confocal microscopy and flow cytometry. Tfn uptake was found to be similar in both Ip6k1 +/+ and Ip6k1 −/− MEFs, suggesting that endocytosis is normal in the absence of IP6K1 ( Figure 1C-E).
Decreased Tfn distribution in the ERC in Ip6k1 −/− MEFs could be due to a delay in Tfn trafficking from endosomes. To determine whether Tfn is held back in endosomes, we stained cells to detect the early endosome marker EEA1 1 h after Tfn uptake (Figure 2A). There is no change in EEA1 levels ( Figure 2B), and it appears distributed throughout the cytoplasm in both cell types (Figure 2A). In Ip6k1 −/− MEFs, there is significant colocalization of endocytosed Tfn with EEA1-positive structures in the cytoplasm, whereas in Ip6k1 +/+ MEFs, the early endosomes lack Tfn (Figure 2A, colocalization panel). To quantify these observations, we determined the percentage of Tfn that colocalizes with EEA1 throughout the cell. Ip6k1 −/− MEFs had significantly higher colocalization of Tfn in EEA1-containing vesicles when compared with Ip6k1 +/+ MEFs (Figure 2A,C). This suggests that Tfn exit from early endosomes towards the ERC is slower in cells lacking IP6K1. Expression of active but not inactive IP6K1 rescued this defect, implying that inositol pyrophosphates are required for endosomal sorting of Tfn (Figure 2A,C).

IP6K1 activity is required to maintain Golgi morphology
Cytoplasmic dynein is required to position the Golgi apparatus in the pericentriolar region. In cells with defective dynein function, the Golgi appears fragmented [29]. When stained for the cis-Golgi marker GM130, most Ip6k1 +/+ MEFs showed a normal perinuclear arc-like appearance of the Golgi, whereas a significant fraction of Ip6k1 −/− MEFs displayed a fragmented Golgi complex ( Figure 3A,B). This is a hallmark phenotype of defective dynein function in mammalian cells. Expression of catalytically active but not inactive IP6K1 was able to correct the Golgi morphology in Ip6k1 −/− MEFs ( Figure 3A,B), suggesting that the synthesis of inositol pyrophosphates is required for Golgi maintenance.

Vesicle movement is slower in cells lacking IP6K1
Defects in motor-dependent trafficking lead to impaired vesicle movement along microtubule tracks. To monitor vesicle motility, we used CT-B, which binds to the cell surface ganglioside GM1, and is endocytosed and trafficked to the trans-Golgi network [30]. We allowed fluorescently labeled CT-B to bind cells on ice for 1 h and noted no difference in CT-B binding to the cell surface in Ip6k1 +/+ and Ip6k1 −/− MEFs ( Figure 4A,B). Next, we allowed cells to endocytose CT-B for 30 min under regular culture conditions, and tracked the movement of fluorescent vesicles. We excluded endocytic vesicles that were close to the plasma membrane and perinuclear vesicles that are likely to have completed their movement towards the trans-Golgi network. The rate of transport of CT-B-containing endosomes is significantly lower in Ip6k1 −/− compared with Ip6k1 +/+ MEFs (Figure 4C,D,G and Supplementary Videos S1 and S2). CT-B vesicle movement was restored upon expression of the catalytically active but not the kinase-dead form of IP6K1 ( Figure 4E-G and Supplementary Videos S3 and S4), confirming that the intracellular levels of inositol pyrophosphates influence vesicle motility.
Dynein-dependent vesicle movement to the cell interior can be assessed by monitoring the intracellular distribution of phagosomes, which are transported along microtubules towards perinuclear lysosomes [31]. We isolated peritoneal macrophages from Ip6k1 +/+ and Ip6k1 −/− mice, allowed them to phagocytose latex beads for 1 h, and observed their distribution 1 h postinternalization. Most of the beads were observed in the perinuclear region in Ip6k1 +/+ macrophages, whereas in Ip6k1 −/− macrophages, a larger number of the beads were away from the nucleus ( Figure 5A). Analysis of the fractional distance of each bead from the nuclear centroid revealed a greater nuclear proximity of phagosomes in Ip6k1 +/+ macrophages compared with Ip6k1 −/− macrophages ( Figure 5B,C). These data also reveal that the dependence of dynein on IP6K1 is conserved in cell types other than fibroblasts.

Dynein-driven endosomal motility is reduced in slime mold lacking inositol pyrophosphates
To study the influence of cellular inositol pyrophosphates on motility at the level of individual vesicles, we turned to the slime mold D. discoideum. Since the cellular functions of dynein in D. discoideum are highly similar to those seen in mammalian cells [32], this slime mold has been extensively used to examine dyneindependent motor activity in vitro [23]. D. discoideum has 200-500-fold higher levels of inositol pyrophosphates compared with mammalian cells [4,18]. Deletion of the IP 6 kinase i6kA in the amoeba results in the absence of any detectable cellular inositol pyrophosphates [18]. We assessed the in vitro motility of endosomes derived from wild-type and i6kA − D. discoideum on polarity-labeled microtubules [23]. As these microtubules are polymerized in vitro using tubulin purified from goat brain, this assay rules out any influence of inositol pyrophosphates on microtubule assembly or its posttranslational modifications. The fraction of motile endosomes (that were a mix of plus-and minus-end moving) was not significantly different in wild-type and i6kA − amoebae (Table 1). However, the percentage of dynein-driven minus-end-directed motile endosomes was substantially lower in i6kA − D. discoideum. The apparent increase in kinesin-dependent plus-end-directed motile endosomes is most likely a consequence of reduced dynein-driven motility, as both classes of motor proteins are recruited onto endosomes and are reported to engage in a mechanical tug-of-war [24]. We observed no change in the velocity of minus-end-or plus-end-directed endosomes, suggesting that single-molecule properties and ATPase activity of both motor proteins are not compromised in i6kA − amoebae. These observations confirm that inositol pyrophosphates regulate dynein-dependent vesicle movement even in lower eukaryotes such as slime mold.

Dynein IC is pyrophosphorylated by 5-IP 7
Our data so far show that lowering cellular levels of inositol pyrophosphates leads to reduced dynein-dependent vesicle transport. The dynein complex contains two large catalytic heavy chains that move the motor along microtubules and several small noncatalytic subunits that function in vesicle attachment and dynein structure maintenance [13]. One mechanism by which inositol pyrophosphates may directly influence dynein function is by serine pyrophosphorylation on dynein subunits. A consensus pyrophosphorylation site comprises one or more Ser residues flanked by Asp/Glu residues [9][10][11]33]. This is also a preferred site for phosphorylation by the protein kinase CK2, which is known to prephosphorylate the target Ser to prime it for pyrophosphorylation [9]. CK2 has been shown to phosphorylate the N-terminus of the dynein IC-2C [34,35], a noncatalytic dynein subunit, which possesses multiple phosphorylation sites, including a pyrophosphorylation consensus sequence (Supplementary Figure S1). The N-terminus of IC-2C, the IC isoform expressed in all tissues [19], is intrinsically disordered [36,37] and can be divided into a charge cluster and a Ser-Pro-rich region (Supplementary Figure S1). To identify the sites of CK2 phosphorylation on the IC-2C N-terminus, we expressed the N-terminal 111 amino acid residues of mouse IC-2C as a fusion to GST and phosphorylated it with CK2 in vitro. Mass spectrometry revealed two sites of CK2 phosphorylation in the charge cluster (Ser46 and Ser51) and one site in the Ser-Pro cluster (Ser98; Figure 6A and Supplementary Figure S2). All three sites have been shown to be endogenously phosphorylated (Supplementary Figure S1). From amongst these three CK2 sites, Ser51 lies amidst Glu and Asp residues and constitutes a preferred site for pyrophosphorylation [9].
To test for IP 7 -mediated pyrophosphorylation, we incubated CK2-prephosphorylated IC fragments, IC(1-70) encompassing the charge cluster, and IC(1-111) which includes the charge and Ser-Pro clusters, with radiolabeled IP 7 . We observed that IC(1-111) undergoes pyrophosphorylation, whereas IC(1-70), which harbors the consensus IP 7 target site, is not pyrophosphorylated ( Figure 6B). Interestingly, mutating the consensus IP 7 site by replacing Ser51 with Ala abolished pyrophosphorylation in IC(1-111) ( Figure 6B), implying that Ser51 is indeed the target of IP 7 . The absence of pyrophosphorylation in the IC(1-70) fragment suggests that the Ser-Pro cluster (residues 71-111) is required to facilitate pyrophosphorylation on Ser51. Interestingly, the site of pyrophosphorylation we have identified in mouse IC-2C is well conserved in human and rat (Supplementary Figure S1), suggesting that the effect of IP 7 on dynein is likely to be conserved in these species. In D. discoideum, the IP 7 target Ser is conserved, but the neighboring Asp and Glu residues are replaced with Thr. These Thr residues may undergo phosphorylation to mimic Asp/Glu and create a consensus site for pyrophosphorylation. To assess whether IC undergoes pyrophosphorylation in vivo, we carried out a 'back-pyrophosphorylation' assay [8,10]. This is currently the accepted method to assess IP 7 -mediated pyrophosphorylation in vivo, as mass spectrometry, the preferred method for detection of phosphorylation, cannot distinguish between pyro-versus bisphosphorylated peptides [38]. As an IP 7 target protein isolated from Ip6k1 +/+ MEFs is already pyrophosphorylated in vivo, it has been shown to display diminished incorporation of radiolabeled phosphate from Beads with a fractional distance of ≤0.4 were classified as 'perinuclear', and each cell was categorized as having either ≥60 or <60% perinuclear beads. Data were analyzed by a two-tailed Fisher's exact test, P ≤ 0.001. Table 1 Motility of endosomes derived from D. discoideum on polarity-labeled microtubules In vitro motility of refractile endosomes in PNS from wild-type and i6kA − D. discoideum was assayed on polarity-labeled microtubules. Data were compiled from two independent experiments. For statistical analysis, the endosome velocities (mean ± SEM) were analyzed by a two-tailed Student's t-test. Contingency tables (2 × 2) of the number of motile vs. nonmotile endosomes and of the number of plus-end-vs. minus-end-directed endosomes were subjected to a two-tailed Fisher's exact test. The number of endosomes used to measure velocity is indicated in brackets. The observed motion of each endosome was parsed into multiple velocity segments by a Bayesian algorithm.

Pyrophosphorylation of IC positively regulates its interaction with p150 Glued
Dynein IC facilitates dynein recruitment to vesicles via its interaction with the p150 Glued subunit of the dynactin complex [40,41]. The first 106 residues in the N-terminal region of IC-2C contain binding sites for the CC1 domain of p150 Glued [41]. To test the effect of IC pyrophosphorylation on its interaction with p150 Glued , phosphorylated and pyrophosphorylated mouse IC(1-111) were assessed for their ability to pull down native p150 Glued from human HEK293T cell lysates. Human and mouse dynein IC-2C show 95% sequence identity in the N-terminal region, and the CC1 region of p150 Glued is 99% identical, suggesting that the regulation of dynein-dynactin interaction is likely to be conserved in these species. IC(1-111) that was CK2-phosphorylated and then pyrophosphorylated by 5-IP 7 shows enhanced interaction with p150 Glued compared with the CK2-phosphorylated fragment alone ( Figure 7A). IP 6 or a nonhydrolyzable analog of 5-IP 7 (5-PCP-IP 5 ) [42], which is likely to bind but cannot pyrophosphorylate the target protein, induces only a marginal enhancement in p150 Glued binding to IC(1-111) ( Figure 7A).
To examine whether the loss of pyrophosphorylation on IC in Ip6k1 −/− MEFs alters its interaction with p150 Glued , we conducted coimmunoprecipitation assays after cross-linking proteins with a reversible cross-linker. We noted a significant decrease in the extent of IC and p150 Glued interaction in extracts from Ip6k1 −/− MEFs compared with Ip6k1 +/+ MEFs ( Figure 7B,C). To monitor if the decrease in IC-p150 Glued interaction leads to reduced dynein recruitment on vesicle membranes, extracts from Ip6k1 +/+ and Ip6k1 −/− MEFs were subjected to differential centrifugation to isolate the membrane fraction. The amounts of IC, p150 Glued , and the Golgi matrix protein GM130 in the membrane fraction were normalized to their levels in the TH. The ratio of IC in the MP, compared with total cell homogenate, was significantly lower in Ip6k1 −/− MEFs compared with Ip6k1 +/+ MEFs ( Figure 7D). In contrast, the ratios of p150 Glued and GM130 were unchanged in the same extracts.
Earlier studies have shown that overexpression of the IC N-terminal region in HeLa cells produces a dominant-negative effect by binding to p150 Glued and preventing its interaction with the endogenous dynein complex, leading to Golgi disruption [41]. Therefore, we hypothesized that overexpression of the IC(1-111) wild-type fragment should disrupt Golgi morphology as it is capable of pyrophosphorylation and dynactin binding, whereas the S51A mutant, which cannot be pyrophosphorylated, should not exhibit Golgi fragmentation due to its reduced binding to p150 Glued . IC(1-111) fragments were cotransfected with the enhanced YFP-tagged trans-Golgi marker β1,4-galactosyltransferase (EYFP-Golgi) [41], and the percentage of Golgi disruption was scored. The cells were categorized into intact Golgi, partially dispersed Golgi if more than three fragments were seen, and completely dispersed Golgi if the EYFP signal was scattered across the cytoplasm ( Figure 7E). Approximately 50% of cells transfected with IC(1-111) showed intact Golgi, whereas 30 and 20% cells exhibited partially dispersed and completely fragmented Golgi, respectively ( Figure 7F). In contrast, overexpression of IC(1-111)S51A was unable to disrupt the Golgi apparatus, as the Golgi distribution pattern in these cells was comparable to the vector control ( Figure 7E,F). These results suggest that pyrophosphorylation of the dynein IC is required for effective binding of IC to p150 Glued inside cells.

Discussion
Our study has identified inositol pyrophosphates as novel regulators of dynein function in protozoan and metazoan cells. Inositol pyrophosphate-mediated serine pyrophosphorylation of the dynein IC promotes its interaction with the p150 Glued subunit of dynactin, suggesting that pyrophosphorylation of IC may regulate attachment of the dynein motor to vesicle membranes. Corroborating this, we observed a decrease in IC-p150 Glued association and decreased dynein binding to membranes in cells lacking IP6K1, leading to multiple defects in dynein-dependent vesicle transport in these cells. In fact, the phenotypic defects in Golgi maintenance and Tfn sorting observed in Ip6k1 −/− MEFs are similar to the phenotypes observed in dynein IC knockdown cells [28]. The dynein and dynactin protein assemblies interact at multiple sites, including a recently elucidated association of the dynein heavy chain with the Arp filament of dynactin, which is stabilized by the cargo adaptor protein Bicaudal-D2 [43,44]. However, the best characterized direct association between the dynein and dynactin complexes is the interaction between the coiled-coil region of p150 Glued and the N-terminus of dynein IC [40,41]. In the absence of dynactin, the N-terminus of rat IC-2C, which is identical with mouse IC-2C (Supplementary Figure S1), is monomeric and disordered [37]. Residues 10-44 lying within this region constitute a core binding site for the coiled-coil CC1 domain of p150 Glued . Two IC monomers bind a CC1 dimer to form a tetrameric complex held together predominantly by electrostatic interactions [37]. Given that disordered sequences are prone to posttranslational modifications, phosphorylation on residues in IC-2C surrounding the core binding site could influence its interaction with p150 Glued [36,45]. Phosphorylation of Ser84 and Thr89 inhibits IC-p150 Glued interaction, whereas phosphorylated Ser81 and Ser83 have no effect, and the role of phosphorylation at other sites (Supplementary Figure S1) is unknown [35,46,47]. Earlier work documenting CK2 phosphorylation of the IC-2C N-terminus did not identify the Ser residues phosphorylated by CK2 and was therefore unable to discern its influence on IC-p150 Glued interaction [34,35]. Here, we identified three Ser residues in the IC-2C N-terminus that are phosphorylated by CK2 ( Figure 6A and Supplementary Figure S2), but observe no effect of CK2 phosphorylation on its interaction with full-length p150 Glued ( Figure 7A). On the other hand, subsequent IP 7 -mediated pyrophosphorylation of Ser51, which lies in close proximity to the core p150 Glued -binding region, significantly enhanced the binding of IC(1-111) to p150 Glued ( Figure 7A). Thus, pyrophosphorylation of dynein IC serves to stabilize IC-p150 Glued association ( Figure 7A-C), switching dynein from an unbound to a membrane-bound state ( Figure 7D). Increasing membrane recruitment of multiple dynein motors would counteract the effect of kinesin, which is known to contest with dynein to regulate the directionality of organelle movement along microtubules [24]. Therefore, IC pyrophosphorylation may act as a regulatory signal to enhance minus-end-directed vesicle motility.
Pyrophosphorylation of the cargo adaptor protein subunit AP3B1 [10] and the yeast glycolytic transcription factor GCR1 [11] inhibits interaction with their respective binding partners, whereas we report augmented p150 Glued binding as a consequence of dynein IC pyrophosphorylation, suggesting that this posttranslational modification can influence protein-protein interaction in diverse ways. Our study also sheds light on the mechanism of protein pyrophosphorylation, revealing that a disordered protein sequence distinct from the target Ser residue is required for pyrophosphorylation. Similarly, pyrophosphorylation target sites identified in three different yeast RNA Pol I subunits also fall within highly mobile regions [33], implying that flexible secondary structures promote phosphotransfer from IP 7 to phosphoserine. As with other posttranslational modifications, pyrophosphorylation has recently been shown to be reversible [48], suggesting that dynein recruitment to vesicles via IC-p150 Glued interaction may be regulatable by yet to be identified signaling pathways.