|Home | About | Journals | Submit | Contact Us | Français|
Membrane motility is a fundamental characteristic of all eukaryotic cells. One of the best-known examples is that of the mammalian Golgi apparatus, where constant inward movement of Golgi membranes results in its characteristic position near the centrosome. While it is clear that the minus-end-directed motor dynein is required for this process, the mechanism and regulation of dynein recruitment to Golgi membranes remains unknown. Here, we show that the Golgi protein golgin160 recruits dynein to Golgi membranes. This recruitment confers centripetal motility to membranes, and is regulated by the GTPase Arf1. Further, during cell division, motor association with membranes is regulated by the dissociation of the receptor-motor complex from membranes. These results identify a cell cycle-regulated membrane receptor for a molecular motor, and suggest a mechanistic basis for achieving the dramatic changes in organelle positioning seen during cell division.
The positioning of the Golgi apparatus near the centrosome-based microtubule-organizing center is a striking and physiologically relevant feature of the interphase mammalian cell. Golgi membranes are constantly captured by outgrowing microtubules and actively translocated towards the microtubule minus ends (Ho et al., 1989; Presley et al., 1997) by the dynein and dynactin motor complex (Burkhardt et al., 1997; King and Schroer, 2000; Roghi and Allan, 1999). Once positioned near the centrosome, Golgi membranes are linked laterally to form the ribbon-like membrane network, after which the Golgi ribbon may be anchored directly to the MTOC (Infante et al., 1999; Takahashi et al., 1999).
Several fundamental processes rely on an accurately positioned Golgi including differentiation of myoblasts into myotubes, immunological synapse formation, neuronal arborization, and directed cell migration. The pericentrosomally positioned Golgi apparatus, in combination with oriented microtubule arrays, defines a physiologically important axis of secretion to the most proximate aspect of the plasma membrane. In response to a polarity cue, cells reorient their microtubule array and reposition the Golgi apparatus towards the stimulus defining the cell leading edge (Kupfer et al., 1982; Pu and Zhao, 2005). Specific disruption of Golgi positioning blocks polarized delivery of secretory cargo to the leading edge, resulting in a collapse of cell polarity and an inability to migrate and heal a wound (Yadav et al., 2009).
Golgi positioning is dramatically regulated during these processes. When myoblasts fuse to form a myotube, the pericentrosomal Golgi ribbon fragments and is repositioned as isolated Golgi stacks that encircle each nucleus (Ralston, 1993). This repositioning may be caused by a loss of Golgi membrane motility, and could be an obligate part of the muscle differentiation pathway (Lu et al., 2001). When natural killer cells and cytotoxic T cells form an immunological synapse with target cells, the Golgi repositions towards the synapse to secrete lytic factors that kill the target cell (Kupfer et al., 1983; Stinchcombe et al., 2006). In hippocampal neurons, the Golgi aligns on the side facing a newly forming axon (de Anda et al., 2005). In pyramidal neurons, the cell body-localized, or somatic, Golgi orients towards the apical dendrites (Horton et al., 2005). Interestingly, neuronal Golgi elements are also present in dendrites as multiple Golgi “outposts”, whose positioning is required for dendritic arborization (Ye et al., 2007).
Mitotic regulation of Golgi positioning is perhaps even more striking. As microtubules reorganize to form the spindle-pole body, the Golgi membrane network fragments and then completely vesiculates giving rise to a mixture of Golgi vesicles and Golgi vesicle clusters that are dispersed throughout the mitotic cytoplasm (Shorter and Warren, 2002). The uncoupling of membranes from their microtubule-based positioning is thought to ensure their uniform partitioning during cell division (Yadav and Linstedt, 2011). At the end of mitosis, the Golgi membranes once again move towards microtubule minus ends and reestablish their pericentrosomal position in each daughter cell.
Despite its importance, our understanding of the mechanisms that position the Golgi and move secretory cargo inward is incomplete. Particularly vexing is that the components that specifically link the dynein/dynactin complex to these membranes to confer pericentrosomal positioning remain unknown (Kardon and Vale, 2009). Although there are several dynein-interacting proteins on the Golgi, none of these have been shown required for membrane association of the motor. Identifying the membrane receptor for the molecular motor is critical, as it will help us understand both the regulation of motor recruitment during membrane transport and organelle positioning, and the regulatory events that allow the motor to switch to its specialized roles during cell division.
Golgin160, a homodimeric coiled-coil protein localized primarily to cis Golgi cisternae (Hicks et al., 2006; Hicks and Machamer, 2002), is an excellent candidate receptor because its depletion blocks Golgi positioning and yields dispersed ministacks (Yadav et al., 2009). Here, unlike all previously identified dynein interacting proteins, we show that golgin160 satisfies a list of stringent criteria expected from a candidate motor receptor on the Golgi. Golgin160 was specifically required for dynein recruitment and Golgi motility. It directly bound dynein and its dynein-binding site was required for Golgi positioning. Golgin160 was also sufficient to confer functional motor recruitment. Golgin160 directly bound the Arf1 GTPase and this interaction was responsible for membrane association of the golgin160 motor complex. Finally, golgin160 dissociated from Golgi membranes at mitosis and this dissociation was required for mitotic Golgi dispersal. Together, our data suggest that the Golgi positioning depends on constant motility provided by the recruitment of dynein by golgin-160, a cell cycle regulated motor receptor. The membrane attachment of this motor receptor, through the highly regulated GTPase Arf1, is likely a key control point for the dramatic changes in motility and organelle positioning observed during cell differentiation, polarization, and division.
We first confirmed the requirement for golgin160 in Golgi positioning by depleting golgin160 with siRNAs individually targeting distinct exons shared by known splice isoforms. Consistent with our previous work, each siRNA yielded loss of pericentrosomal Golgi positioning (Figure S1) suggesting a loss of minus end motility. To assay Golgi motility directly, it was assessed after washout of the microtubule depolymerizing agent nocodazole in cells co-expressing the fluorescently tagged Golgi protein mCherry-GRASP55 and the microtubule plus-end protein EB1-GFP using high-resolution live cell confocal microscopy. In control cells, microtubule outgrowths were coupled to inward tracking movements of the Golgi membranes resulting in a pericentrosomally positioned Golgi complex (Figure1A upper panels and Movie S1). Consistent with previous work (Blum et al., 2000; Ladinsky et al., 1999; Presley et al., 1997; Simpson et al., 2006), inward-tracking membranes appeared tubular presumably due to distension of the membrane by force exerted by the motor. In contrast, in cells lacking golgin160, although outward microtubule growth was normal, there was no inward tracking of Golgi membranes (Figure 1A lower panels, 1B and Movie S2). These cells failed to establish a pericentrosomal Golgi. Unlike controls, even when Golgi objects were grazed by outward tracking plus-ends, there was no inward Golgi movement in golgin160-depleted cells (Figure 1C). Thus, Golgi membrane inward motility specifically depends on golgin160.
Golgin160 was also critical for centripetal motility in the dynein driven process of ER-to-Golgi transport. In control cells, release from the restrictive temperature initiated ER-to-Golgi and then Golgi-to-plasma membrane trafficking of temperature sensitive viral cargo protein VSVG-GFP (Figure 1D, upper panels, 0-12 min and 12-20 min, respectively). Tracking movements were exclusively inward during the fist period and outward during the second period. In golgin160 depleted cells, however, VSVG-GFP moved in a non-directed fashion as it emerged from the ER in the first time period, but then tracked persistently outward as it left the Golgi in the second time period (Figure 1D, lower panels). Directional persistence of movement during the ER-to-Golgi time period was significantly inhibited in these cells, whereas that during the Golgi-to-PM time period was not (Figure 1E).
To test whether Golgi membranes, upon golgin160 depletion, lacked minus-end motility because dynein was not recruited, antibodies against Tctex1, a subunit of the dynein motor complex, were used to localize the motor (Tai et al., 1998). Depletion of Tctex1 confirmed antibody specificity (Figure S2A). Control cells showed dynein on Golgi membranes, while cells lacking golgin160 did not (Figure 2A). Dynein remained evident on nocodazole-induced Golgi ministacks, indicating that neither Golgi fragmentation nor loss of microtubules displaced the motor on their own. Line profiles showed that Golgi objects coincided with peaks of dynein fluorescence except in the absence of golgin160 (Figure 2B). Total dynein fluorescence on Golgi objects decreased by more than 90% in cells lacking golgin160, compared to control cells (Figure 2C). The golgin160-dependence of dynein Golgi localization was confirmed using an antibody against dynein heavy chain (Figure S2B) and this effect was specific because other peripheral membrane proteins (GMAP210, GBF1 and golgin97) remained Golgi-localized in cells depleted of golgin160 (Figure S2C). Consistent with these microscopic assays, dynein recovery on Golgi membranes isolated using anti-giantin-coated magnetic beads was reduced by 70% by golgin160 knockdown, whereas control proteins remained membrane associated (Figure 2D). Thus, depletion of golgin160 inhibited recruitment of dynein to Golgi membranes.
To elucidate the mechanism of golgin160 mediated dynein recruitment, we first tested whether golgin160 binds dynein. Indeed, the dynein intermediate chain (DIC) and golgin160 co-immunoprecipitated using either anti-golgin160 or anti-DIC antibodies (Figure 3A). The quantified recovery of golgin160 in anti-DIC immunoprecipitates was 10.3±0.7 % (n=8). The dynactin subunit p150 was also present in anti-golgin160 precipitates, indicating that the golgin160-dynein complexes contained the dynein regulatory complex dynactin (Schroer, 2004). Control golgins, GM130 and giantin, did not precipitate with the anti-DIC antibody (Figure 3A).
The golgin160 dynein-binding site was then mapped. GFP-tagged constructs were generated (Figure 3B) corresponding to the golgin160 N-terminal domain (Nterm), which contains its Golgi localization determinant (Hicks and Machamer, 2002), and the remaining C-terminal section enriched in coiled-coil segments (cc2-8). Whereas Nterm failed to bind dynein, cc2-8 yielded a robust interaction (Figure 3C). Further dissection showed that the seventh coiled-coil segment (cc7) was required and sufficient for dynein binding (Figure 3C). The cellular localization of these constructs strongly supported the interaction assays. Note that, at the expression levels used, none of the constructs interfered with the Golgi localization of endogenous golgin160 (not shown). As expected, full-length and Nterm localized to Golgi membranes (Figure 3D). However, constructs that lacked the Golgi localization region but contained the dynein interaction region cc7, including cc7 itself, localized to the cytoplasm and centrosomes (Figure 3D). Dynein moves towards, and is localized at, centrosomes (Roghi and Allan, 1999) so partial centrosome localization is expected for a binding partner of dynein. In contrast, the construct cc2-6 that lacked cc7 was cytoplasmic (Figure 3D). Further, endogenous golgin160 was centrosome localized after being displaced from the Golgi by BFA and centrosome localization of both endogenous golgin160 and cc7 was nocodazole sensitive (Figure S3). Thus, centrosome localization was a result of dynein mediated motility rather than binding to a centrosome core component.
These results prompted us to test whether cc7 directly binds dynein and whether this binding is functionally significant. The dynein intermediate chain (DIC) is known to mediate several key dynein interactions (Kardon and Vale, 2009; King et al., 2003) via its N-terminus while its C-terminus binds the dynein heavy chain. Consistent with this, purified His-DIC1-237 bound purified GST-cc7 of golgin160 (Figure 4A). His-DIC1-237 did not bind GST alone or GST-cc8. Binding of His-DIC1-237 to GST-cc7 was saturable with an apparent Kd of 1.2 μM (Figure 4B). Thus, the golgin160 cc7 domain directly interacts with the dynein motor via the DIC N-terminus.
If this interaction is functionally significant, then a version of golgin160 lacking cc7 should fail to rescue the golgin160 knockdown phenotype. Indeed, whereas expressing siRNA-resistant, full-length golgin160 in cells depleted of endogenous golgin160 restored pericentrosomal Golgi positioning, a golgin160 replacement construct lacking cc7 did not (Figure 4C-D). The replacement construct appeared stable and properly localized to the Golgi indicating that the cc7 dynein-binding domain is required for Golgi positioning.
As dynein recruitment by golgin160 was required for Golgi positioning, we examined whether this was sufficient to confer pericentrosomal positioning of a distinct organelle, mitochondria, which is not normally positioned there. Golgin160 was targeted to the outer membrane of mitochondria using the transmembrane mitochondrial-targeting determinant of Tom20 (T20) (Sengupta et al., 2009). Localization to the mitochondria was confirmed by colocalization with Mitotracker (Figure S4A). In contrast to a control construct, GFP-tagged T20, which failed to recruit dynein and left mitochondria dispersed, the chimeric full-length golgin160 construct, T20-G160, induced dynein recruitment to mitochondria and their juxtanuclear clustering (Figure 5). When targeted to mitochondria, the dynein binding domain cc7 was sufficient to cause dynein recruitment and clustering while cc2-6 had neither activity (Figure 5). In radial profile plots, full-length and cc7 showed a peak near the fluorescence centroid (blue lines) indicating strong clustering. This reflected microtubule dependent motility of mitochondria, as the golgin160 induced mitochondrial clustering was blocked by nocodazole (red lines). Indeed, centrosome staining indicated that clustering occurred at the minus ends of microtubules (Figure S4B). Also, clustering was unaffected by Golgi dispersal using BFA indicating that it did not depend on interaction with Golgi membranes (Figure S4C). As a further control, microtubule-independent clustering of mitochondria induced by T20-GM130, which recruits the tether GRASP65 to form cross bridges between mitochondria (Sengupta et al., 2009) was insensitive to nocodazole and did not involve dynein recruitment (Figure 5).
We next investigated the mechanism by which golgin160/dynein complex was targeted to the Golgi. Because golgin160 was displaced from the Golgi by BFA before a Golgi marker redistributed to the ER (Figure S5A), consistent with previous work (Hicks and Machamer, 2002; Misumi et al., 1997), we hypothesized that GTP-bound Arf1 may recruit golgin160 to impart minus end motility to Golgi membranes. Indeed, depletion of the Arf1 guanine nucleotide exchange factor GBF1 displaced golgin160 from the Golgi and blocked Golgi positioning (Figure S5B). To ask whether Arf1 directly regulates the N-terminal Golgi localization domain of golgin160, we first used rapid-acquisition, live cell, confocal imaging to compare the rates of BFA-induced membrane dissociation of Nterm and Arf1. Consistent with direct control of Nterm by Arf1, dissociation of each followed single-phase exponential decay kinetics immediately after BFA addition, with Arf1 (t1/2=27 s) slightly faster than Nterm (t1/2=39 s) (Figure 6A-B). In further support of this, expression of GTP-restricted Arf1-Q71L increased membrane localization of endogenous golgin160 and also conferred BFA resistance. In contrast, inhibiting Arf1 activation by expressing GDP-restricted Arf1-T31N caused membrane dissociation of golgin160 (Figure S5C-D).
Next, we tested whether Nterm directly binds Arf1. GST-Nterm and previously described His-tagged Arf1 constructs (Rein et al., 2002) were purified (Figure 6C). GST-Nterm bound GTP-restricted Arf1-Q71L and wildtype Arf1, but did not bind GDP-restricted Arf1-T31N (Figure 6D) indicating that golgin160 is an Arf1 effector. As a control, GST-cc8 bound none of the His-tagged Arf1 constructs. To determine if membrane association of golgin160 was dependent on the GTPase activity of Arf1, we analyzed the fluorescence recovery of Nterm after photobleaching the Golgi pool. Compared to wildtype Arf1, expression of Arf1-Q71L dramatically impaired the recovery of Nterm after photobleaching on the Golgi confirming that the GTP hydrolysis cycle of Arf1 actively regulates membrane binding of the Golgi localization domain of golgin160 (Figure 6E-F).
Based on our identification of golgin160 as the dynein Golgi receptor we examined whether mitotic Golgi dispersal might be mediated by mitotic inhibition of either the dynein/golgin160 interaction or the golgin160/membrane interaction. Significantly, golgin160 was absent from mitotic Golgi membranes and was now evident on each spindle pole (Figure 7A), implying that during mitosis golgin160 membrane binding was inhibited while its dynein binding was maintained. Line profiles across Golgi objects confirmed the loss of golgin160 fluorescence during mitosis (Figure 7B). Additionally, when cell homogenates were fractionated, mitotic golgin160 was mostly recovered in the cytosolic supernatant fraction (Figure 7C). Consistent with its spindle localization, golgin160 from mitotic cell extracts co-immunoprecipitated with DIC (Figure 7D). Thus, dissociation of golgin160/dynein complexes from membranes provides a straightforward mechanism of mitotic Golgi dispersal.
Next we tested whether golgin160 release from membranes is required for mitotic Golgi dispersal. To bypass the mitotic membrane regulation, we permanently anchored the dynein-binding cc7 domain of golgin160 using the transmembrane domain of the Golgi protein giantin. Significantly, whereas a control membrane-anchored construct localized to mitotic Golgi clusters and had no effect on mitotic Golgi dispersal, the Golgi-anchored cc7 domain induced clustering of mitotic Golgi membranes at each spindle pole (Figure 7E). To quantify this effect, the position of the metaphase plate was used to orient hemi-circles encompassing the Golgi fluorescence in one half of each mitotic cell and the integrated fluorescence intensity for each degree radian was plotted. The peak of fluorescence evident at 90°, which was the axis perpendicular to the metaphase plate, corresponds to the accumulation of Golgi membranes at spindle poles (Figure 7F). Thus, mitotic Golgi membranes failed to disperse when membrane dissociation of golgin160 was specifically prevented. In summary, these experiments indicate that golgin160 dissociates from the Golgi at mitosis and that this dissociation is required for Golgi dispersal.
This work identifies golgin160 as the regulated Golgi-localized receptor for dynein. Golgin160 recruits dynein using its cc7 segment by directly binding dynein through its intermediate chain. This interaction is required and sufficient to confer dynein recruitment, minus-end motility, and pericentrosomal positioning. Golgin160 in turn is localized to Golgi membranes by direct, nucleotide-dependent binding of its N-terminus to Arf1. Together, these observations imply that golgin160 provides an elongated linkage of the motor to Golgi membranes that is regulated by the switch-like behavior of a GTPase (Figure 7G). Further, golgin160 is displaced from mitotic Golgi membranes revealing the mechanism of uncoupling dynein from Golgi membranes to achieve mitotic Golgi dispersal.
A normal Golgi ribbon may be disrupted under a variety of conditions. The disruption of specific proteins by siRNA-mediated knockdown, confirmed by rescue, has provided distinct categories of phenotypes corresponding to distinct steps in Golgi ribbon assembly (Feinstein and Linstedt, 2008; Mukhopadhyay et al., 2010; Puthenveedu et al., 2006; Puthenveedu and Linstedt, 2004; Sengupta et al., 2009; Yadav et al., 2009). Of these, the golgin160 category of dispersed, secretion-competent, Golgi ministacks is clearly the best match for a specific defect in inward motility of Golgi membranes. Because the Golgi apparatus is highly dynamic and its integrity depends on many pathways, stringent tests are needed to establish a candidate as a direct mediator of Golgi motility. The candidate must be: 1) localized on Golgi membranes, 2) required for Golgi localization of dynein 3) required for Golgi positioning and motility, 4) a direct binding partner of the motor complex, 5) sufficient to confer dynein recruitment and motility, and 6) regulated during mitosis. There are other dynein-interacting proteins on the Golgi but only golgin160 meets these criteria. Of particular significance, only golgin160 is known to be required for Golgi association of the dynein motor complex. Most of the previously identified candidates act via dynactin (Vallee et al., 2012) but dynactin knockdown leaves dynein Golgi localized (Haghnia et al., 2007). Accumulating evidence indicates that dynactin is a complicated regulator of dynein activity rather than a specific mediator of motor-cargo binding (Vallee et al., 2012). Bicaudal-D (BICD), a conserved protein that interacts with dynein and dynactin (Hoogenraad et al., 2001; Hoogenraad et al., 2003; Suter et al., 1989), is a well-known candidate but evidence also argues against its role as the Golgi dynein receptor. First, although BICD localizes to the trans Golgi network it is also present on cytoplasmic vesicles, centrosomes, nuclear pore complexes and, in Drosophila, lipid droplets and mRNP particles (Akhmanova and Hammer, 2010; Splinter et al., 2010). Second, overexpression of a BICD construct that lacks the dynein-binding domain affects retrograde Golgi-to-ER trafficking but not Golgi positioning (Matanis et al., 2002). Third, depletion of BICD does not cause fragmentation of the Golgi apparatus (Fumoto et al., 2006). Rather, it perturbs the localization of the centrosomal protein ninein. Intriguingly, some dynein-interacting proteins play specialized, rather than ubiquitous, roles. Rhodopsin interacts with dynein and mediates the movement of post-Golgi, rhodopsin-containing vesicles to the outer segments of photoreceptor neurons (Tai et al., 1999). Lava lamp, a Drosophila protein, binds dynein, and mediates apical movement of Golgi membranes during cellularization of Drosophila embryos (Papoulas et al., 2005). However, it is not known if it is required for dynein recruitment, and, more importantly, it does not have a mammalian homologue.
Golgi positioning is largely the result of microtubule organization and the interaction of Golgi membranes with the dynein motor complex. En route to the Golgi apparatus, membranes that bud from the ER coalesce and fuse forming the ERGIC. Recruitment of the Arf guanine nucleotide exchange factor GBF1 begins the process by which these membranes acquire Golgi-like properties (Garcia-Mata and Sztul, 2003). GBF1 recruits and activates Arf1 and other Arf isoforms. The activated Arf proteins interact with multiple effectors to initiate ER retrieval and ERGIC-to-Golgi trafficking (Szul et al., 2007; Volpicelli-Daley et al., 2005). Our results suggest that golgin160-mediated dynein recruitment plays a role at multiple steps in establishing Golgi positioning. First, golgin160 might act as a critical Arf1 effector, directly recruiting dynein, and therefore imparting inward motility to peripheral ERGIC membranes. This model, that Arf1 recruits dynein through golgin160, is consistent with the findings that supplementing an in vitro reaction with a constitutively activated form of Arf1 increases dynein recovery in a Golgi membrane fraction, and that the GTPase cdc42 negatively regulates Arf1 to control dynein membrane association (Chen et al., 2005; Hehnly et al., 2010). Second, continued presence of Arf1, and therefore golgin160, on Golgi membranes ensures retention of membranes at the MTOC by constant dynein-driven inward motility. Third, Arf1 recruitment of golgin160/dynein complexes to Golgi membranes might also mediate Golgi organization involving Golgi-nucleated microtubules. Golgi microtubules contribute to maintenance of Golgi positioning near the cell center and, upon nocodazole washout, clustering of Golgi membranes in the cell periphery (Efimov et al., 2007; Miller et al., 2009). That these events depend on golgin160 is suggested by the lack of either Golgi cell-center positioning or clustering in the periphery in cells depleted of golgin160.
Arf1-based, nucleotide-dependent, golgin160/dynein recruitment provides a potential model for regulating the cycling of the motor complex on and off Golgi membranes. Spatial control of Arf activity, by relative enrichment of GBF1 activity in the cell periphery and Arf1-GAP activity on centrally localized Golgi membranes could drive a cycle of the motor complex involving cargo capture in the periphery and release in the cell center. Our identification of the receptor as being a peripheral, rather than integral, component of the Golgi membrane suggests that at least a fraction of the motor returns to the cell periphery by diffusion in the cytosol. Whether a significant fraction also returns on retrograde cycling membranes is an interesting question that remains to be answered. Arf-based recruitment of dynein, through its receptor golgin160, suggests that regulation of motor recruitment by GTPases to organelles is a common theme in intracellular trafficking.
As mentioned above, Golgi positioning can change dramatically during diverse physiological scenarios such as differentiation, polarization, immune synapse formation, myotube formation in muscle, and axonal and dendritic outgrowth in neurons. In most cases, the change can be understood in terms of two associated mechanisms: a rearrangement of the microtubule array, and regulation of motor association with the Golgi membranes. The latter, which had been difficult to study due to the dearth of bona fide motor receptors, can now be studied with a focus on dynein/golgin160/Arf1 associations and, especially, the control of Arf1 activation and inactivation. Further work on golgin160 is likely to reveal mechanistic insights into motor regulation in diverse organellar processes.
Mitotic Golgi dispersal occurs in a stepwise fashion. The Golgi ribbon begins to fragment and disperse in late G2, and, as the cell progresses to metaphase, these fragments vesiculate resulting in dispersed vesicles and vesicle clusters. While it has been proposed that inhibition of motility plays a key part in mitotic Golgi disassembly, the mechanisms have been unclear. Our findings that golgin160/dynein complexes are released from Golgi membranes during mitosis, and that this release is required for normal mitotic Golgi dispersal, provides significant mechanistic insight into how motility might be inhibited in mitosis. One obvious possibility is that golgin160 membrane dissociation is mediated by mitotic inhibition of Arf1. Indeed, expression of GTP-restricted Arf1 impairs mitotic Golgi dispersal (Altan-Bonnet et al., 2003) and GBF1 is mitotically phosphorylated and released from membranes, suggesting that it no longer activates Arf1 (Morohashi et al., 2010). Nevertheless, whether Arf1 is inhibited during mitosis is controversial. GBF1 is not the only exchange factor for Arf1, and Arf1 remains active in mitotic extracts (Xiang et al., 2007). It may be that GBF1 function is specific to the pool of Arf1 on ERGIC and cis Golgi membranes. If so, only Arf1 effectors on these membranes, such as golgin160, would be mitotically inhibited. Golgin160 is also regulated by phosphorylation (Cha et al., 2004). Although mitotic phosphorylation of golgin160 has not been studied in detail, another possibility is that golgin160 binding to Arf1 is regulated by phosphorylation of golgin160. Interestingly, under mitotic conditions, golgin160 remained bound to dynein. This is somewhat surprising, as phosphorylation of dynein during mitosis modulates its binding to other partners (Whyte et al., 2008). Whether the golgin160 bound dynein pool has a specific mitotic function is unknown, but it is a strong possibility that release of dynein from the Golgi membranes allows the motor to switch from its interphase function in organelle positioning to its mitotic roles.
In summary, we have identified golgin160 as a key cell cycle regulated dynein receptor that drives Golgi positioning. It is specifically Golgi localized through the interaction of its N-terminus with the GTPase Arf1. It is required for both dynein recruitment and inward motility of Golgi membranes. It directly binds the intermediate chain of dynein. It confers both dynein recruitment and minus-end motility at an exogenous site. And finally, it dissociates from mitotic Golgi membranes to allow Golgi dispersal. The relationship between its role as a dynein receptor and others identified for golgin160, including trafficking of specific cargos (Hicks et al., 2006; Williams et al., 2006), apoptosis (Maag et al., 2005; Mancini et al., 2000), and spermatogenesis (Banu et al., 2002; Matsukuma et al., 1999) is at present unclear. Intriguingly, different splice isoforms of the protein exhibit different cargo binding activities (Hicks and Machamer, 2005). Whether the dynein-binding activity of golgin160 is specific to certain isoforms remains to be tested. Regardless, the present work elucidates a key aspect of organelle positioning and its regulation for accurate organelle partitioning during cell division. It also impacts our understanding of the spatial organization of the secretory pathway, and the processes that depend on this spatial organization such as cell polarity, cell migration and wound healing.
EB1-GFP, pRSET-DIC(1-257), and pTREArf1 were gifts from Drs. Shaw (UCSF), Schroer (JHU) and Vilardaga (U Pitt), respectively. mCh-GRASP55, VSVG-GFP and GFP-golgin160 have been described (Feinstein and Linstedt, 2008; Yadav et al., 2009). Mutagenesis was by Quickchange. Residue numbers of golgin160 constructs were: Nterm (1-496), cc2-8(393-1498), cc2-6(393-1244), cc7(1245-1385). The GFP-golgin160 siRNA resistant construct was made by silent base pair changes (3598GTGGAAGCCGGG to GTCGAGGCCGGC) and residues 1245-1385 were deleted using loopout primers (Sengupta et al., 2009) to generate Δcc7. The Tom-20 fragment (Sengupta et al., 2009) was inserted by PCR to create an N-terminal tag. mCh-Nterm was in pcS2-mCh. GST-N term, GST-cc7 and GST-cc8 were in pGEX4T1 at the SalI and NotI cloning sites. Arf1-GFP, Q71L and T31N were in pEGFP-N3. His-Δ17Arf1 was made using loopout primers in pRSET and Q71L and T31N were then introduced. C-terminally anchored GFP-Gtn and cc7-GFP-Gtn were by appending giantin sequence (residues 3155-3255) at the XbaI site in the corresponding pcS2-GFP vectors. The siRNAs were as follows: control (GACCAGCCATCGTAGTACTTT), golgin160 (exon 2: AACCTGCAACCAAAACGAGAC, exon 11: AGAGAAGTTAAGAGAAGAATT, exon 15: GGCCCTCGCGGCCAAGGAGGC, and exon19: CTCCTTGGAGCTGAGTGAGGT), GBF-1 (Szul et al., 2007) (AAACGAAATGCCCGATGGAGC) and Tctex1 (Palmer et al., 2009) (AAGUGAACCAGUGGACCACAA). Affinity purified rabbit antibodies against Tctex1 were a gift from Dr. Sung (Cornell). Rabbit antibodies against golgin160 were generated against amino acids 1400-1498 (Covance). Antibodies against GPP130 and giantin have been described (Mukhopadhyay et al., 2010). Commercial antibodies were: dynein intermediate chain (Millipore), His (Bethyl labs), GMAP210 and GBF1 (BD Transductions), golgin97 (Invitrogen), γ-tubulin, α-tubulin, and dynein heavy chain (Sigma). Nocodazole (used at 0.5μg/ml for 2 h), brefeldin-A (used at 2.5μg/ml), GDP and GTPγS were from Sigma. Transfection reagents were oligofectamine (Invitrogen) for siRNA and JETpei (PolyPlus) for plasmids and were used according to manufacturer’s instructions.
Live-cell imaging was performed using an Andor Revolution XD Spinning disk system on a Nikon Eclipse Ti inverted microscope with a 100× 1.49 NA objective (Nikon). Solid-state lasers 488nm and 561nm (Coherent) were used for both imaging and photobleaching. Cells were imaged in Opti-MEM (Gibco) with 10% serum and 35 mM HEPES (pH 7.4), maintained at 37°C in a temperature controlled chamber (In Vivo Scientific). Time-lapse images were acquired with an iXon+897 EM-CCD 16-bit camera driven by iQ (Andor). Fixed cell image-stacks were acquired using a Zeiss Axiovert 200 microscope with a 100× 1.4 NA objective (Zeiss) attached to an UltraView spinning-disk confocal system (Perkin Elmer) equipped with three-line laser, independent excitation and emission filter wheels (PerkinElmer) and a 12-bit Orca ER digital camera (Hamamatsu Photonics).
HeLa cells were transfected with siRNA, and then 48 h later were co-transfected with EB1-GFP and mCh-GRASP55 plasmids. After 24 h, the cells were incubated with nocodazole for 2 h and washed with cold media. Within 5 min single optical section 16 bit confocal images were recorded for each channel every 2 sec. Using ImageJ, EB1-GFP comet tracks and mCh-GRASP55 Golgi tracks were analyzed by projecting frames from a period of 20 sec and 100 sec respectively, such that the first frame was color-coded blue, intervening frames were green and the last frame was red. The blue/green/red tracks indicate directionality and persistence of movements (Jaulin and Kreitzer, 2010). All discernable EB1-GFP tracks were counted for five 20 sec periods in a given experiment and the percent of these that showed a blue/green/red order projecting to the cell periphery was determined. For GRASP55-mCh, all discernable tracks showing blue/green/red order projecting inward were counted during five 100 sec periods in a given experiment.
HeLa cells transfected with siRNA were re-transfected 48 h later with VSVG-GFP and mCh-GRASP55 plasmids and after 12 h were shifted to 40°C for 12 h. The cells were mounted and confocal image stacks were acquired every 5 sec at 37°C. ImageJ plugin MTrackJ was used to track VSVG-GFP object movement after max-value projection of the image stack, create projected tracks, and calculate distance at each step. Directional motility was taken to be the ratio of the minimum XY distance between starting and final coordinates over the actual distance traversed. Only objects that tracked through at least 4 successive frames were included. At least 20 tracks per experiment were included for each track cluster.
ImageJ RGB Profiler plugin was used on single optical sections for profile plots. For Golgi level determinations, average value projections were created from confocal image stacks. For Tctex1 levels, Golgi pixels were selected using thresholding of giantin staining and the mean Tctex1 intensity over these pixels was obtained using the “Measure” function of ImageJ. Golgin160 levels on the Golgi were similarly obtained after selecting Arf1-positive pixels except that mean values were multiplied by the area of the selected regions and the results are presented as a percent of total golgin160 fluorescence in each cell, which was determined in the same way after selecting the entire cell.
Anti-Rabbit IgG Dynabeads M-280 (50μl packed, Invitrogen) were incubated with rabbit anti-giantin (20 μl serum) for 30 min at 4°C and then blocked with 0.1% bovine serum albumin in phosphate buffer saline. Postnuclear supernatants were prepared from 150μl of packed HeLa cells 72 h post-transfection and homogenized in an equal amount of buffer (50mM NaCl, 1mM EDTA, 10mM triethanolamine pH7.4) using a 25 g needle and centrifuged 5 min at 1000×g. After a 2 h, 4°C incubation of the postnuclear supernatant with the beads, the beads were collected using a magnet, washed twice and analyzed by immunoblotting with enhanced chemiluminescence and an LAS-3000 imager with ImageGauge software (Fujifilm).
HeLa cells (10 cm plate) were collected, lysed in 200μl buffer (10mM Hepes pH7.2, 50mM KCl, 0.5% Triton-X100, 2mM DTT, 1mM EDTA and protease inhibitors), passed through a 25-gauge needle and incubated for 30 min on ice. The lysate was centrifuged in a microfuge for 5 min. The supernatant was pre-cleared for 30 min using 10 μl packed sepharose beads (Invitrogen) and then incubated overnight with antibody-bound sepharose beads (4 μl anti-DIC with 10 μl Protein G beads or 2 μl anti-golgin160 with 10μl Protein A beads). The bound fraction was treated to three washes with buffer, two washes with buffer lacking detergent, SDS-PAGE and immunoblotting.
GST- and His-tagged proteins were purified as described (Sengupta et al., 2009) and nucleotide was exchanged in the Arf1 proteins (Rein et al., 2002). For DIC binding to golgin160, 2-10 μg of each GST protein was bound to glutathione beads and incubated with His-DIC (1.2 μM) for 1 h at 4°C in 200ul buffer (10 mM Hepes pH7.2, 50mM KCl, 0.5% triton, 2mM DTT, 1mM EDTA and protease inhibitors). After 3 washes with buffer and 2 washes with buffer lacking detergent, the bound fraction was analyzed by Ponceau S staining and immunoblotting. For Kd determination, the assay was carried out using 2.3 μM GST protein and 0-6.8 μM His-DIC and the resulting DIC immunoblot signal was quantified using ImageJ and Prism (GraphPad Software Inc., La Jolla, CA) by non-linear regression analysis assuming one-site binding. For Arf1 binding to golgin160, 2-10 μg of each GST protein was bound to glutathione beads and incubated with the nucleotide-exchanged Arf1 constructs at 1.2 μM for 1 h at 4°C in binding buffer (25mM Hepes pH6.8, 300mM potassium acetate, 1mM DTT, 0.5mM MgCl2, 0.2% triton and 0.7mM GDP or GTPγS). The beads were washed twice with the binding buffer and twice with the binding buffer without detergent (Rein et al., 2002). Anti-His immunoblot or Ponceau S staining, respectively, determined Arf1 and GST protein level.
Cells were transfected with siRNA using oligofectamine (Invitrogen) and then after 48 h the cells were transfected with GFP or siRNA-resistant GFP-tagged golgin160 constructs. After another 24 h, the cells were fixed using paraformaldehyde and immunostained for giantin. The percent rescue was the percent of GFP-positive cells showing a juxtanuclear Golgi normalized by the knockdown penetrance (~90%), which was the fraction of knockdown cells showing dispersed Golgi after expression of GFP alone (Puthenveedu and Linstedt, 2004).
After 24 h, transfected cells were left untreated or treated for 3 h with nocodazole and immunostained using anti-DIC antibody. Quantification of clustering was performed on average value projections of confocal image stacks using the ImageJ plugin Radial profile as described (Sengupta et al., 2009). Grayscale thresholds were set for each experiment and the centroid of the fluorescence for each cell was determined using the “Measure” function. The normalized fluorescence intensity in each concentric circle drawn from the determined centroid was obtained using the Radial profile plug-in.
To compare Golgi dissociation kinetics of golgin160 and Arf1 to GPP130, HeLa cells were transfected with mCh-GPP130 and either Arf1-GFP or GFP-Nterm and after 24 h were imaged at 3 frames/sec in two channels before and after BFA (2.5μg/ml) addition. The fluorescence in the Golgi region over time was determined using the ImageJ “Measure” function as a ratio of the initial Golgi fluorescence. Prism5 was used to fit that data to a single-phase exponential decay equation to calculate the t1/2. For fluorescence recovery after photobleaching, HeLa cells were transfected with mCh-Nterm and either Arf1-GFP or Q71L-GFP. After 24 h, a pre-bleach image was acquired and the Golgi region was selected and bleached using 488 and 568 nm laser light. Confocal image stacks were then acquired every 5 s for each channel. The fluorescence in the Golgi region over time was determined as a function of the initial Golgi fluorescence and Prism5 was used to fit that data to a single-phase association equation using Prism5.
HeLa cells stably expressing GalNAcT2 were fixed in paraformaldehyde and stained with anti-golgin160 and Hoechst. Metaphase cells were identified based on presence of a metaphase plate. RGB Profile plots were drawn using ImageJ. Mitotic cell extracts were obtained from HeLa cells treated with 0.5 μg/ml nocodazole for 12-16 followed by shake-off. Co-immunoprecipitation was as above. For fractionation, the cells were homogenized in buffer (250mM sucrose, 10mM HEPES pH7.2, 1mM EDTA, 1mM DTT and protease inhibitors) using a 25-gauge needle and centrifuged at 1000×g for 5 min. The postnuclear supernatant was then centrifuged at 100,000×g for 60 min to yield the cytosol fraction, which was precipitated using trichloroacetic acid and washed with acetone. The pellet fraction was briefly rinsed and both the cytosol and membrane pellet fractions were analyzed by immunoblotting. For analysis of cells expressing membrane anchored GFP-Gtn and cc7-GFP-Gtn constructs, cells were fixed 24 h post-transfection and stained using anti-tubulin and Hoechst. To assay Golgi distribution the ImageJ Azimuthal Average plugin was used. Mitotic cells in thresholded, average-projected, image stacks were divided into hemispheres along the metaphase plate. Each hemisphere was then selected and using the Azimuthal Average plugin, divided into sectors of one degree radian and the normalized Golgi fluorescence intensity in each degree sector was measured.
+ Golgin160 is the dynein membrane receptor for Golgi positioning.
+ The GTPase Arf1 controls membrane association of golgin160.
+ Golgi dispersal at mitosis is mediated by membrane dissociation of golgin160.
We thank T.H. Lee, S. Mukhopadhyay, D. Sengupta, C.Bachert and C. Priddy for critical reading of the manuscript and T. Schroer (JHU) and C. Sung (Cornell) for essential contributions. Study supported by NIH grant GM056779 to A.D.L.
Author Information The authors declare no competing interests.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.