Regulation of vesicle transport and cell motility by Golgi-localized Dbs

Abstract

DBS/MCF2L has been recently identified as a risk locus for osteoarthritis. It encodes a guanine nucleotide exchange factor (Dbs) that has been shown to regulate both normal and tumor cell motility. In the current study, we have determined that endogenous Dbs is predominantly expressed as 2 isoforms, a 130 kDa form (Dbs-130) that is localized to the Golgi complex, and an 80 kDa form (Dbs-80) that is localized to the endoplasmic reticulum (ER). We have previously described an inhibitor that binds to the RhoGEF domain of Dbs and blocks its transforming activity. Here we show that the inhibitor localizes to the Golgi, where it specifically interacts with Dbs-130. Inhibition of endogenous Dbs-130 activity is associated with reduced levels of activated Cdc42, enlarged Golgi, and resistance to Brefeldin A-mediated Golgi dispersal, suggesting a role for Dbs in vesicle transport. Cells treated with the inhibitor exhibit normal protein transport from the ER to the Golgi, but are defective in transport from the Golgi to the plasma membrane. Inhibition of Dbs-130 in MDA-MB-231 human breast tumor cells limits motility in both transwell and wound healing assays, but appears to have no effect on the organization of the microtubule cytoskeleton. The reduced motility is associated with a failure to reorient the Golgi toward the leading edge. This is consistent with the Golgi localization, and suggests that the Dbs-130 regulates aspects of the secretory pathway that are required to support cell polarization during directed migration.

Abbreviations

BFA

Brefeldin A

COP1

coat protein

DAPI

4’, 6-diamidino-2-phenylindole

Dbs

Dbl's big sister

DH

Dbl homology

ECL

electrochemiluminescence

ER

endoplasmic reticulum

FACS

fluorescence-activated cell sorting

FGD1

faciogenital dysplasia 1 protein

GEF

guanine nucleotide exchange factor

GFP

green fluorescent protein

HA

hemagglutinin

HM

homogenization medium

MTOC

microtubule organizing center

NF-1

neurofibromatosis type 1

PAK3

p21 protein-activated kinase 3

PBS

phosphate buffered saline

PH

pleckstrin homology

rtPCR

real-time polymerase chain reaction

SH3

Src homology 3

siRNA

small inhibitory RNA

VSVG

vesicular stomatitis virus-G

WGA

wheat germ agglutinin

Introduction

Dbs/MCF2L is a member of the RhoGEF family that was independently identified by 2 groups in screens for cDNAs whose expression cause deregulated proliferation in fibroblasts.^1,2 More recently, MCF2L was identified as a risk locus for osteoarthritis in a genome-wide association scan.³ Like most RhoGEF family members, Dbs contains a RhoGEF domain module that can catalyze the exchange of GDP for GTP on Rho family substrates.⁴ We and others have shown that Dbs has in vitro catalytic activity specific for RhoA,^2,5 Cdc42,^2,5 and, to a lesser extent, RhoG.⁶ In addition to its RhoGEF domain, full-length Dbs encodes a COOH-terminal Src homology 3 domain (SH3), 2 spectrin-like repeats, and a recently identified NH₂-terminal Sec14p domain (Fig. 1A). Although the native function of Dbs is not known, overexpressed Dbs is predominantly found in a perinuclear region where it colocalizes with a marker for the Golgi.⁷ It is widely expressed during mouse development⁸ and in adult mouse and rat tissue, with highest expression in the brain.^1,2,8 Expression has also been reported in hematopoietic cells of the myeloid lineage,¹ Schwann cells,⁹ endothelial cells isolated from normal colonic mucosa (ref. 10; reported as the KIAA0362 protein), and in normal and tumor-derived breast epithelial cells.¹¹ Although Dbs is predominantly expressed as a 4.2 kb transcript in mice¹ (5.0 kb in rat²), smaller transcripts are also expressed in a tissue-specific manner,^1,2 and several splice variants have been identified in cDNA libraries and validated by rtPCR.⁸ These variants all contain the central RhoGEF domain but differ in the presence or absence of the N-terminal Sec14 domain and the COOH-terminal SH3 domain. Whether or not these alternative messages are translated to produce functional protein has not yet been determined.

Figure 1.

Identification of 130 kDa and 80 kDa variants of Dbs that localize to the Golgi apparatus and endoplasmic reticulum respectively. (A) Schematic representation of the protein domain structure of full-length human Dbs. In addition to its RhoGEF domain (DH/PH), Dbs encodes a COOH-terminal Src homology 3 domain (SH3), 2 spectrin-like repeats (Spec), and a recently identified NH₂-terminal Sec14p domain (Sec14). (B) Two isoforms of Dbs are expressed in HeLa cells. Cells were transfected with control siRNAs or siRNAs directed at Dbs. Lysates were collected at 48 h and examined by western blot for expression of Dbs and tubulin (loading control). (C) Distribution of endogenous Dbs. HeLa cells were examined by co-immunofluorescence with an antibody for endogenous Dbs and for the Golgi (GM130). In the merge on the right panel, the nucleus is also stained with DAPI. (C, D) Localization of the 130 kDa and 80 kDa variants of Dbs to the cytoplasmic face of the Golgi and ER respectively. Intact ER (D) or ER/Golgi (E) organelle fractions were isolated from HeLa cells. (C) The ER fraction was immunoprecipitated with a Dbs antibody or IgG control, and then examined by western blot with an antibody for calreticulin, Dbs, or GM130 (left panels). In addition, whole cell lysates were immunoprecipitated with a Dbs antibody and then examined by western blot for the presence of calreticulin to confirm that calreticulin and Dbs do not directly interact (right panels). (D) The ER/Golgi organelle fraction was immunoprecipitated with a GM130 antibody (to isolate purified Golgi) or IgG control, and then examined by western blot for the presence of Dbs, GM130, and calreticulin (left panels). Also, whole cell lysates were immunoprecipitated with Dbs and GM130 antibodies and then examined by western blot for the presence of GM130, Dbs and calreticulin to confirm that GM130 and Dbs do not directly interact (right panels).

Several recent studies suggest an important role for Dbs in both normal and tumor cell motility.^9,11 It had been previously shown that neurotrophin-3 acts through the TrkC receptor tyrosine kinase to promote the migration of premyelinating Schwann cells.^12,13 This pathway is mediated by both Cdc42 and Rac1, and Yamauchi et al. recently showed that Dbs links TrkC to Cdc42 (but not Rac) activation.⁹ The association between overexpressed Dbs and TrkC does not occur at the plasma membrane, but rather within a perinuclear region and within intracellular punctate structures. In a related study, Dbs was shown to regulate internalization of the transferrin receptor in a Rac-dependent manner.¹⁴ Again Dbs-induced Rac activation was localized to the perinuclear region suggesting that it is the pool of Dbs that is resident within this subcellular compartment that can regulate vesicle trafficking and cell motility. In a more recent study, the role of Dbs in cell motility was confirmed in tumor-derived breast epithelial cells.¹¹ Thus, overexpression of full-length human Dbs is sufficient to induce Cdc42-dependent motility in T47D cells, while siRNA-mediated silencing of endogenous Dbs limits the Cdc42-dependent motility of MDA-MB-231 cells.

Whereas overexpressed full-length Dbs is found primarily in the perinuclear region, oncogenic Dbs is found at the plasma membrane.⁷ As with other RhoGEF family members, the transforming version of Dbs contains a truncation that is responsible for oncogenic activation,¹ and in a recent study, we have mapped this autoinhibitory activity to the NH₂-terminal Sec14p domain.⁷ The Sec14p protein was first identified and characterized in S. cerevisiae as a phosphatidylinositol transfer protein that was required for transport of secretory proteins from the Golgi complex.¹⁵ In vitro, Sec14p mediates the transfer of phosphatidylinositol and phosphatidylcholine between membrane bilayers.¹⁶ The Sec14p structure consists of 2 domains: an N-terminal α-helical domain which is required for lipid transfer, and a C-terminal lipid interacting domain.¹⁷ Although there are no structural homologs of Sec14p in mammalian cells, regions corresponding to the C-terminal lipid-binding domain of Sec14p have been described in the context of mammalian Rho regulatory proteins including GTPase Activating Proteins and RhoGEFs.¹⁸ In Dbs, this domain was found to be responsible for phosphoinositide-dependant sequestration of Dbs to the perinuclear region.⁷ Since both the Sec14 and PH domains of Dbs have been shown to bind phosphoinositides, these domains presumably cooperate to regulate the subcellular distribution and positioning of Dbs on specific membrane compartments.

In addition to regulating cellular distribution, the Sec14 domain of Dbs can also bind directly to and inhibit the isolated RhoGEF domain of Dbs.⁷ Based on this observation, we designed a Dbs inhibitor that contains a fragment of the Sec14 domain (Dbs-Sec14). We showed that this inhibitor can block the transforming activity of onco-Dbs, but not other RhoGEF family members, if expressed in trans. In the current study, we have used a human version of this inhibitor to examine the endogenous function of Dbs. We have shown that the inhibitor specifically targets an isoform of Dbs that is localized to the Golgi. This isoform utilizes Cdc42 as a substrate and regulates vesicle transport from the Golgi. By fusing the inhibitor with GFP, we are able to show that Dbs regulates components of the secretory pathway that are required to establish cell polarity during directed cell migration.

Results

Endogenous Dbs is localized to the Golgi complex and endoplasmic reticulum

In order to explore the endogenous function of Dbs, we first examined its expression in HeLa cells. Initially, cell lysates were collected and examined by western blot using a Dbs-specific antibody (Fig. 1B). This antibody recognizes 2 predominant bands at 130 kDa and 80 kDa, which was confirmed using a second commercially available antibody (not shown). The larger band corresponds closely with the predicted molecular weight of full-length Dbs. Both bands were significantly reduced when the cells were transfected with Dbs-specific siRNAs suggesting that both represent Dbs isoforms. In order to examine the cellular distribution of Dbs, immunofluorescence was performed (Fig. 1C). A strong concentration of Dbs is observed in the perinuclear region which is consistent with the distribution of overexpressed Dbs in NIH 3T3 cells.⁷ The distribution of endogenous Dbs is reminiscent of Golgi and/or ER staining, and co-localization with a Golgi-specific marker (GM130) is observed (Fig. 1C, right panel). Localization of endogenous Dbs to the Golgi was confirmed using a second commercially available antibody (not shown). To determine if the 2 Dbs isoforms localize to discrete membrane compartments, we performed whole organelle immunoprecipitations. Intact ER was isolated and then immunoprecipitated using Dyna-beads coated with a Dbs antibody. This antibody immunoprecipitated an isoform of Dbs, found to be approximately 80 kDa, along with the ER marker calreticulin (Fig. 1D, left panel). Co-immunoprecipitation of whole cell lysates shows that the association of the 80 kDa isoform of Dbs with calreticulin is not direct (Fig. 1D, right panel). We then isolated an organelle fraction that contained the Golgi and ER. Since these membrane compartments could not be completely separated by centrifugation, we immunoprecipitated the Golgi from the mix using the GM130 antibody (Fig. 1E). This antibody immunoprecipitated organelles that contain the 130 kDa variant of Dbs and GM130, but not the 80 kDa variant or calreticulin. Co-immunoprecipitation of whole cell lysates revealed that the association of GM130 with the 130 kDa isoform of Dbs is also not direct (Fig. 1E, right panel). Collectively, our observations suggest that the 80 kDa variant of Dbs is localized to the the ER while the 130 kDa variant is localized to the the Golgi.

The Sec14 domain of Dbs localizes predominantly to the Golgi

We have previously shown that the localization of overexpressed murine Dbs to the perinuclear region of NIH 3T3 cells is mediated by the N-terminal Sec14 domain. Overexpressed variants of Dbs that lack the Sec14 domain are more broadly distributed within the cell with prominent staining at the plasma membrane. Thus, we determined whether the Sec14 domain of human Dbs is sufficient for Golgi localization. For comparison, we also examined the isolated Sec14 domains of Dbl and NF-1. Dbl is a second RhoGEF family member that has a similar domain structure to Dbs. Using immunofluorescence we observed that when expressed alone, the isolated Sec14 domain of Dbs colocalizes precisely with a Golgi marker (Fig. 2A). In contrast, the isolated Sec14 domain of Dbl has a more diffuse distribution throughout the cytoplasm, while the NF-1 domain is widely distributed in a more punctate pattern. To confirm that the isolated Sec14 domain of Dbs is excluded from the ER, we performed organelle fractionations (Fig. 2B). Two fractions were obtained, a fraction that contained ER (Calreticulin but no GM130) and a fraction that contained both ER and Golgi (Calreticulin and GM130). Whereas the isolated Sec14 domain localized to the ER/Golgi fraction, it did not localize to the ER fraction alone.

Figure 2.

The isolated Sec14 domain of DBS localizes to the Golgi. (A) Cells were transfected with the indicated FLAG-tagged constructs and then examined by immunofluorescence for cellular distribution of the overexpressed proteins. Co-localization of the indicated Sec14 domains with the Golgi was demonstrated using the WGA marker. (B) Sub-cellular fractionation of cells transfected with the Dbs Sec14 domain or cognate vector (pAX). Fractions were western blotted for the overexpressed protein, and both Golgi and ER markers (GM130 and calreticulin).

The isolated Sec14 domain binds endogenous Dbs and limits Cdc42 activation

We have previously shown that a fragment of the Sec14 domain of murine Dbs can bind to the isolated RhoGEF domain, and inhibit transformation if both constructs are expressed in trans. Thus, we determined whether this fragment can be used to inhibit endogenous, Golgi-localized Dbs. Initially, we performed co-immunoprecipitations to confirm that the human Dbs Sec14 domain interacts with its RhoGEF domain. For this analysis, we co-expressed the isolated FLAG-tagged Sec14 domain with an HA-tagged RhoGEF domain (DH/PH). We were readily able to detect an interaction between the Sec14 domain and the RhoGEF domain whether we performed the immunoprecipitation using an anti-FLAG antibody (Fig. 3A), or an anti-HA antibody (Fig. 3B). To further map the interaction we also performed co-immunoprecipitations using fragments that contain the isolated DH domain and the isolated PH domain. Although both fragments were expressed at equivalent levels, we observed that an interaction is only observed with the catalytic DH domain. Next, we determined whether the Sec14 domain could interact with endogenous full-length Dbs. For this analysis, we overexpressed the Dbs Sec14 domain alone in HeLa cells. As an additional control we expressed the isolated Sec14 domain of Dbl, also in HeLa cells (Fig. 3C, lower panel). As shown in Figure 3C, we were readily able to coimmunoprecipitate the 130 kDa and the 80 kDa (not shown) isoforms of Dbs from cell lysates with the Dbs Sec14 domain, but not endogenous Dbl. This suggests that both Dbs isoforms contain the RhoGEF domain, and that the exclusion of the inhibitor from the ER is not attributable to its inability to interact with the 80 KDa isoform. Interestingly, we were unable to coimmunoprecipitate endogenous Dbl or Dbs with the Dbl Sec14 domain. Since we have shown previously that endogenous Dbs utilizes Cdc42 as a substrate, we determined whether the Dbs Sec14 domain can affect Cdc42-GTP levels. Thus, we expressed the Dbs Sec14 domain in HeLa cells and performed affinity precipitation assays (Fig. 3D and E). We consistently observed a significant 25% reduction in Cdc42-GTP levels in cells expressing the Sec14 domain which suggests that endogenous, Golgi-localized Dbs is utilizing Cdc42 as a substrate, and that the Sec14 domain can be used as an inhibitor of this interaction.

Figure 3.

The isolated Sec14 domain interacts with endogenous Dbs and limits Cdc42 activation. (A and B) Cells were co-transfected with the indicated combinations of constructs encoding the HA-tagged RhoGEF domain of Dbs (DHPH), the HA-tagged PH domain of Dbs (PH), the HA-tagged DH domain of Dbs and the FLAG-tagged Sec14 domain of Dbs (Sec14). Lysates were collected at 48 h after transfection and examined by western blot (IB) for expression of the Dbs fragments using an anti-HA antibody (lower panel) (A), or an anti-FLAG antibody (arrow: lower panel) (B). IP indicates antibody used for immunoprecipitation while IB indicates an antibody used for a western blot. (C) Cells were transfected with the isolated FLAG-tagged Sec14 domain of Dbs, Dbl, or cognate vector (pAX) as indicated. Lysates were collected at 48 h and examined by western blot for expression of endogenous Dbl (lower blot). IP indicates antibody used for immunoprecipitation while IB indicates an antibody used for a western blot. Arrows indicate the position of the FLAG-tagged Dbs and Dbl Sec14 domains. (D) Cells were transfected with the isolated FLAG-tagged Sec14 domain of Dbs or with cognate vector (pAX). Lysates were collected at 48 h and examined by western blot for expression of endogenous Cdc42 (total) and overexpressed Sec14 (FLAG). Lysates were then subjected to affinity purification with GST-PAK3. GTP-bound Cdc42 (active) was visualized by western blot. (E) Quantification of band intensity was done using the Odyssey infrared imager (Li-Cor). Quantification results are depicted as the ratio of intensity of activated Cdc42 to total Cdc42 averaged from 3 independent experiments.

Dbs mediates BFA-dependent Golgi dispersal

To explore a role for Dbs in Golgi dynamics, cells were transfected with the Dbs inhibitor in the absence or presence of BFA. As an additional control, cells were also treated with BFA in the presence of the isolated Dbl Sec14 domain. BFA treatment completely disrupted the Golgi in over 95% of cells transfected with vector alone (Fig. 4A and D) or with the isolated Sec14 domain of Dbl (Fig. 4B and D). However, in cells that express the Dbs inhibitor, the Golgi was rendered almost completely insensitive to BFA treatment (Fig. 4C and D). Since BFA is an inhibitor of secretion that is thought to work by promoting the fusion of the cis Golgi with the ER, and the trans Golgi with the endosome,¹⁹ these results suggest that Dbs may be required to support secretory and/or fusion events associated with these 2 membrane compartments. In support of this possibility, we observed that the Golgi in cells that express the Dbs inhibitor are consistently enlarged relative to adjacent non-expressing cells or those transfected with cognate vector. To confirm this observation, cells were transfected with the inhibitor and then Image ProPlus software was used to compare the size of the Golgi in cells that express the inhibitor compared with adjacent cells that don't (Fig. 4E and F). This analysis revealed that on average, expression of the inhibitor leads to a 4-fold increase in Golgi size. This increase in Golgi size was not associated with a significant increase in overall cell size.

Figure 4.

Dbs mediates BFA-dependent Golgi dispersal. Cells were transfected with cognate vector (A), the Sec14 domain of Dbl (B), or the Sec14 domain of Dbs (A). At 24 h post-transfection, cells were treated with Brefeldin A (BFA) for 30 min, and then fixed and stained with a Golgi-specific marker (WGA), DAPI and the FLAG antibody. Cells expressing the Sec14 domain constructs (>50) were then scored for the presence or absence of an intact Golgi. D, data are represented as the percentage of cells with intact Golgi and are averaged for 3 independent experiments. (E) False color image showing an enlarged Golgi in a cell overexpressing the Dbs Sec14 domain. Arrow indicates Golgi while dotted line indicates cell cortex. (F) Quantification of Golgi size from Dbs Sec14 expressing cells and adjacent non-expressing cells.

Dbs regulates the secretory pathway

In order to directly test the possibility that endogenous Dbs regulates the secretory pathway we utilized the GFP-VSVG reporter construct derived from the ts045 mutant strain of vesicular stomatitis virus. When expressed in HeLa cells at 40°C the VSVG-GFP fluorescence is localized exclusively to the ER (Fig. 5A). When the temperature is shifted to 32°C the protein redistributes to the Golgi complex within 20 min, which is then followed by a redistribution of the marker to the plasma membrane (distinguishable at 90 min). In cells co-transfected with the Dbs inhibitor, we observe that the marker redistributes normally to the Golgi membrane after 20 min (Fig. 5A and B), but the number of cells in which the marker is redistributed to the plasma membrane at 90 min drops by approximately 40% (Fig. 5A and C). This suggests that ER-Golgi transport can proceed normally in the presence of the inhibitor, but the secretory pathway from the Golgi to the plasma membrane is impaired.

Figure 5.

Dbs regulates the secretory pathway. Cells were transfected with VSVG-GFP along with the Dbs inhibitor or cognate vector. At 24 h post transfection cells were shifted to 40°C for 12 h to trap the VSVG protein in the ER. Cells were then shifted to 32 °C for 20 min (middle panels) or 90 min (lower panels), and then fixed and stained for expression of VSVG-GFP and the inhibitor. (A) Upper panels show representative images of VSVG-GFP trapped in the ER at 40°C. Middle panels show representative images indicating that the Dbs inhibitor does not prevent movement of the VSVG-GFP protein to the Golgi. Lower right panels show a cell that expresses the inhibitor in which no VSVG-GFP protein is found at the plasma membrane. Arrows indicate membrane accumulation in control cells. Quantification of VSVG transport was performed by manually counting the number of cells expressing VSVG for both the vector and the inhibitor, which have Golgi localization at 20 min (B) or plasma membrane localization at 90 min (C).

Golgi-localized Dbs supports breast cancer cell motility in transwell assays

We have previously shown that Dbs is expressed in the highly motile MDA-MB-231 human, breast tumor cell line, and that Dbs activity is required to support cell motility. Thus, we determined whether Golgi-localized Dbs is required to support the motile phenotype. Initially we performed immunofluorescence to determine the localization of endogenous Dbs in these cells (Fig. 6A). Consistent with the localization in HeLa cells, we observed that Dbs is predominantly localized in a perinuclear region in MDA-MB-231 cells, where it co-localizes with a marker for the Golgi (GM130). In addition, we observed a punctate staining pattern within the cytoplasm with the heaviest distribution of particles between the Golgi and the leading edge. To determine whether Golgi-localized Dbs contributes to cell motility, we transfected cells with either GFP alone, or a GFP fusion of the Dbs inhibitor and performed a wound healing assay (Fig. 6B). Cells expressing the GFP protein were readily able to migrate into an open wound after 19 h. However, when cells expressed the Dbs Sec14-GFP fusion protein motility was significantly impaired (Fig. 6B). The effect of the inhibitor on motility was also tested in a transwell assay (Fig. 6C). Cells were transfected with the indicated constructs and then FACS sorted based on GFP expression. Using this approach, we observed that cells that express the inhibitor show an approximately 35% reduction in motility relative to cells that express GFP alone. Since this is similar to what we previously observed using siRNAs that target Dbs,¹¹ we conclude that the Golgi-localized Dbs protein is the primary pool of endogenous Dbs that controls cell motility.

Figure 6.

Golgi-localized Dbs supports breast cancer cell motility (A) Distribution of endogenous Dbs in MDA-MB-231 cells. Cells were examined by immunofluorescence to determine the cellular distribution of Dbs. Cells are co-stained with a Dbs antibody, a GM130 antibody, and DAPI. (B) MDA-MB-231 cells were transiently transfected with the GFP-Sec14 domain of Dbs, or GFP alone, and then examined in a wound healing assay. The number of GFP positive cells that migrated into the wound was calculated at 19 h and expressed as a percentage of total GFP positive cells at the wound edge (B). (C) MDA-MB-231 cells were transiently transfected with the GFP-Sec14 domain of Dbs, or GFP alone and then sorted by GFP and examined in a transwell motility assay on filters pre-coated with type I collagen. The motility of cells expressing Dbs Sec14 domain is expressed relative to that of cells expressing cognate vector. Nine fields of view were scored from 3 independent transwells for both transfected groups of cells. Results are expressed as the average number of cells counted in each field of view ± SD of 3 independent experiments performed in triplicate.

Dbs supports Golgi-reorientation during directed migration

Despite the significant reduction in the number of cells that traversed the filter in transwell assays, many cells were clearly still able to move in the presence of the inhibitor. Thus, we explored the possibility that Dbs may be required for directed migration, but not random movement. Since directed motility in mammalian cells typically requires repositioning of the Golgi toward the leading edge²⁰ we determined whether impaired motility was associated with a failure to orient the Golgi toward the wound (Fig. 7). For this analysis, HeLa and MDA-MB-231 cells were transfected with the inhibitor, or cognate vector. Cultures were wounded and then allowed to reorient for 3 h. Cells were then fixed, and stained with a marker for Golgi (WGA), actin (phalloidin), and the overexpressed inhibitor (Fig. 7A and B). In control cells, approximately one third of the cells are orientated toward the Golgi prior to serum stimulation which represents the expected random probability (the horizontal lines in Figure 7C represent this baseline). Three hours after wounding, the percentage of control cells in which the Golgi is orientated toward the wound increases to approximately 60% in both cell types. This reorientation is substantially reduced if Dbs expression is suppressed by the Sec14 inhibitor. Since Golgi reorientation is dependent upon the integrity of the microtubule cytoskeleton, we determined whether the Dbs inhibitor disrupts the microtubule network in the same cell type. Thus, cells were transfected with the Dbs inhibitor and then fixed and stained with a tubulin antibody. Using this approach, we were unable to detect any difference in the microtubule network or the overall cell morphology in the presence of the inhibitor (not shown).

Figure 7.

Dbs supports Golgi-reorientation during directed migration HeLa and MDA-MB-231 cells were examined for serum-dependent Golgi reorientation in a wound healing assay as described in materials and methods. Confluent cells were transfected with the Dbs inhibitor or cognate vector overnight and then wounded. After wounding, cells were cultured in 10% serum for 3 h and then fixed and stained for Sec14 domain, Golgi marker (WGA), actin marker (Phalloidin), and DAPI (blue). (A, B) Greater than 50 cells at the wound edge were then scored as Golgi orientated or not orientated. Representative HeLa and MDA-MB-231 cells are shown. Arrows indicate the direction of movement of the cells toward the wound. (C) Data are represented as the average fraction of oriented cells from 3 independent experiments ± SD. Horizontal lines indicate the fraction predicted to be orientated by random chance.

Discussion

Dbs was originally identified based on its ability to induce focus formation in NIH 3T3 cells,¹ and subsequent studies identified RhoA as the relevant substrate for this transforming activity.²¹ This oncogenic derivative of Dbs was expressed from a truncated cDNA that lacked both the N-terminal Sec14 domain and the C-terminal SH3 domain. Unlike onco-Dbs, which is mislocalized to the plasma membrane, overexpressed proto-Dbs accumulates in the perinuclear region and lacks focus-forming activity.⁷ In the current study, we have examined the distribution of endogenous Dbs in human cells. We have identified 130 kDa and 80 kDa isoforms of Dbs and demonstrated by whole organelle immunoprecipitation that they localize to the Golgi and ER, respectively. Although the structural relationship between these 2 variants is unclear, our observation that the isolated Sec14 domain localizes to the Golgi suggest that the 2 variants may differ in the presence of this domain.

Although the Sec14p protein was first identified as a Golgi protein in S. cerevisiae,¹⁵ and we have observed that the Sec14 domain of Dbs also localizes to the Golgi, it does not appear that all mammalian Sec14 domains can function as Golgi anchors. Thus, although Dbs and Dbl share a common structural organization, and have over 42% amino acid identity over their length, their isolated Sec14 domains have different cellular distributions. Unlike the Sec14 domain of Dbs, the Sec14 domain of Dbl is widely distributed in the cell and has a more diffuse staining pattern. When we examine the cellular distribution of the Sec14 domain of the NF-1 protein,²² which is unrelated to Dbl or Dbs, we observe a punctate distribution throughout the cytoplasm. Thus, although we cannot rule out the possibility that all Sec14 domains function as membrane anchors, it seems clear that the endomembranes that they associate with can be distinct.

The difference between the Sec14 domains of Dbs and Dbl also extends to the autoinhibitory function. We have previously shown that the Sec14 domain of Dbs forms a direct, autoinhibitory interaction with its RhoGEF domain.⁷ Using this information, we designed a Dbs inhibitor based on a fragment of the Sec14 domain, and showed that this inhibitor can block the transforming activity of onco-Dbs, but not another RhoGEF family member, Lsc.⁷ In the current study, we have constructed a human version of the inhibitor and shown that it can interact with endogenous Dbs, but not endogenous Dbl. Since endogenous levels of activated Cdc42 are reduced by this inhibitor, it appears to target the RhoGEF function of Dbs in the Golgi. Our observation that levels of active Cdc42 are reduced by an average of 25% suggests that Golgi-localized Dbs is an important, but not exclusive, activator of Cdc42 in these cells. Unlike Dbs, previous studies have demonstrated that the autoinhibitory domain of Dbl lies outside the Sec14 domain,²³ and in the current study we cannot detect an interaction between the Dbl Sec14 domain and either endogenous Dbs or Dbl. Conversely, the Sec14 domain of Dbs does not interact with endogenous Dbl. This suggests that the inhibitory interaction between the inhibitor and endogenous Dbs is specific, and restricted to the Golgi compartment.

Using the Dbs inhibitor we have been able to demonstrate a role for Dbs in Golgi-mediated vesicle transport. Initially, we demonstrated that the inhibitor completely blocks BFA-mediated fragmentation of the Golgi apparatus. Although the precise molecular mechanism through which BFA disrupts the Golgi is unknown, it is widely accepted that it interferes with the retrograde transport pathway from the Golgi to the ER.¹⁹ Thus, the primary target of BFA is a subset of the Sec7-type exchange factors that utilize Arf1p as their substrate.²⁴ Activated Arf1 recruits COP1 proteins to membranes of the cis Golgi where it facilitates the formation of transport vesicles that are targeted to the ER.²⁵ This pathway is also regulated by Cdc42,²⁶ which also binds to COP1 proteins,²⁷ and is a substrate for Dbs in the Golgi. It has been proposed that BFA interferes with this retrograde pathway which leads to fusion of cis-Golgi membranes with the ER, and trans-Golgi membranes with the endosome.¹⁹ Since the activity of Dbs is required to support BFA-mediated membrane redistribution, it is likely to participate in this retrograde pathway.

Further support for a function for Dbs in Golgi-mediated vesicle transport comes from our observation that post-Golgi transport of cargo to the plasma membrane is impaired by the Dbs inhibitor. Although our studies suggest that secretory proteins can reach the Golgi from the ER, they become entrapped within this compartment suggesting transport within the Golgi, or from the Golgi to the plasma membrane is impaired. These results are consistent with a recent study suggesting that Dbs regulates transferrin receptor endocytosis in HeLa cells.¹⁴ In this study, the authors observed that overexpressed Dbs localizes to the perinuclear region where it inhibits receptor endocytosis. Based on the perinuclear localization, it was proposed that Dbs may be involved in the recycling of proteins involved in endocytosis to the plasma membrane. The Cdc42-specific RhoGEF, FGD1, has also been localized to the trans-golgi network where it regulates secretory activity in a Cdc42-dependent manner.²⁸ Loss of endogenous FGD1 activity in HeLa cells has a similar phenotype to loss of Dbs function suggesting that both exchange factors may regulate discrete aspects of vesicle transport within the Golgi compartment. If in fact Dbs supports transport from the Golgi to both the ER and the plasma membrane, this would provide an explanation for the increased Golgi size that we observe in HeLa cells that express the inhibitor.

Our observation that Dbs can regulate vesicle transport from the Golgi also provides an explanation for its contribution to cell motility.²⁹ We have previously shown that endogenous Dbs supports the motility of human breast tumor cells,¹¹ and others have shown an equivalent function in Schwann cells.⁹ Although there are a number of cellular processes that are required to support the motile phenotype, overexpression of full-length Dbs, is sufficient to induce motility in tumor-derived breast epithelial cells suggesting that it may play an important regulatory role in cell movement.¹¹ In the current study, we have identified the Golgi-localized isoform of Dbs as the critical regulator of cell movement. Thus, the loss of cell motility associated with the Golgi-targeted inhibitor, is consistent with the loss of motility associated with Dbs-specific siRNAs.¹¹

Based on our current study, we propose that Dbs supports the motile phenotype by regulating cell polarity. During directed migration, necessary features of the polarized, motile cell include the reorientation of the microtubule organizing center (MTOC) and Golgi complex toward the direction of migration.²⁰ The reorientation of the Golgi and microtubules is thought to facilitate the trafficking of the protein and lipid components that would be required at the leading edge for membrane remodeling and protrusion to occur, and for the formation and turnover of adhesion complexes.^30,31 Our observation that the Dbs inhibitor blocks serum-dependent Golgi re-orientation in wound healing assays, implicates Dbs in the establishment of cell polarity. The reorientation and stabilization of the Golgi toward the leading edge in motile mammalian cells involves a complex interplay between the secretory pathway and the microtubule network. The microtubule cytoskeleton and the Golgi are coordinately polarized toward the leading edge in migrating cells, and structural and regulatory interactions between the microtubule and actin networks are required to sustain this polarity.^32–35 Although the Dbs inhibitor does not dramatically alter the microtubule cytoskeleton in HeLa cells or MDA-MB-231 cells, it is possible that Dbs regulates the targeting of cargo to the leading edge which is required to support the localized capture and/or stabilization of microtubules. In this regard, it is worth noting that although inhibition of Dbs limits directed motility, many cells that express the inhibitor are still able to migrate into the wound or transverse collagen-coated filters. This is reminiscent of studies using inhibitors of microtubule polymerization which demonstrate that, although microtubules are required for proper positioning of the leading edge in motile cells, random motility can still occur.^32–35

Materials and Methods

Molecular constructs

The mammalian expression vector pAX142 has been described previously,¹ while the pEGFP-N1 vector was purchased from Clontech. All pAX142 and pEGFP-N1-based constructs were generated by PCR-based cloning and validated by sequencing. The cDNA for human DBS (a.k.a. MCF2L, OST, ARHGEF14) was obtained from the Invitrogen Mammalian Gene Collection (4138009). cDNAs containing the human Dbs and NF-1 Sec14 domains were obtained from Dr Richard Cerione and Dr Andre Bernards, respectively. pAX-FLAG-DBS and pAX-FLAG-DBS-Sec14 encode full-length (residues 1–984) and the isolated Sec14 domain (residues 1–240) of Dbs, respectively, fused at the N-terminus to a FLAG tag. pAX-DBS-HA-DH/PH, pAX-DBS-HA-DH, and pAX-DBS-HA-PH encode the RhoGEF (residues 575–946), DH (residues 575–797), and PH (residues 785–946) domains of Dbs fused at the N-terminus to a hemagglutinin (HA) tag. pAX-FLAG-NF1-Sec14 and pAX-FLAG-DBL-Sec14 encode the Sec14 domains of NF-1 (residues 1573–1726) and Dbl (residues 1–110), respectively, fused at the N-terminus to a FLAG tag. pEGFP-DBS-Sec14 encodes the Dbs Sec14 domain (residues 1–240) fused at the carboxyl terminus to GFP. The GFP-VSVG reporter construct derived from the ts045 mutant strain of vesicular stomatitis virus was obtained from Dr Jennifer Lippincott-Schwartz.

Cell Culture and Transfection

293T, MDA-MB-231, and HeLa cells were maintained at 37°C (5% CO₂) in Dulbecco's Modified Eagle's medium supplemented with 10% fetal bovine serum (Benchmark, 100–106), 100 unit/ml penicillin, and 100 ug/ml streptomycin (Gemini, 400–109). Cells were transfected using Lipofectamine 2000 reagent (Invitrogen, 11668–019) following the manufacturer's recommended protocol.

Immunoprecipitation, western blot analysis, and immunostaining

Immunoprecipitation and western blot analyses were performed as previously described.⁷ The following antibodies were used for western blots: anti-FLAG (Sigma, M2), anti-HA (Santa Cruz), anti-Ost (Santa Cruz, sc-901), anti-Dbs (Santa Cruz, sc-26125), anti-Cdc42 (Santa Cruz, sc-840), anti-GM130 (BD Transduction Laboratories, 610823), and anti-Calreticulin (Abcam, ab14234). Western blots were imaged using the Odyssey infrared imager (Li-Cor) or ECL reagent (Santa Cruz Biotechnology, sc-2048). Immunostaining of overexpressed or endogenous protein was performed as described previously³⁶ using the following antibodies: anti-FLAG (Sigma, M2), anti-GM130 (BD Transduction Laboratories, 610823), anti-Dbs (Santa Cruz, sc-26125), and anti-Ost (Santa Cruz, sc-901). The Golgi apparatus was stained using Alexa Fluor 488 conjugate of wheat germ agglutinin (Molecular Probes, W-11261). Actin was stained with Alexa 647 phalloidin (Invitrogen, A22287). The nuclei were stained using DAPI (Calbiochem, 268298).

Sub-cellular fractionation

Isolation of organelles by sub-cellular fractionation using a discontinuous sucrose gradient was performed as previously described by others.³⁷ Approximately 10⁸ HeLa cells were transfected with the pAX-FLAG-Sec14 or cognate vector. After 24 h, cells were collected, washed 2 times in homogenization medium (HM, 0.25 M sucrose, 10 mM Tris-Cl, pH 7.4), and re-suspended in 1 mL of HM with 1 ul/mL protease inhibitor cocktail set III (Calbiochem). The cells were homogenized using ∼15 strokes of a Dounce homogenizer. The homogenate was then made 1.4 M by addition of 2 volumes of 2M sucrose. The discontinuous sucrose gradient was established by gently layering decreasingly dense sucrose solutions in a centrifuge tube. The gradient was composed of the following sucrose solutions from bottom to top: 1.6 M, 1.4 M homogenate, 1.2 M, 0.8 M. Each sucrose solution contained 10mM Tris-Cl, pH 7.4 with 1 ul/mL protease inhibitor. The discontinuous gradient was then centrifuged for 2 h at 110 000 × g at 4°C using a swinging bucket rotor. The Golgi/E.R. fraction was obtained from the 0.8 M/1. Two M interface, while the purified E.R. fraction was found at the 1.2 M/1.4 M interface. Isolated organelles were boiled in 2X SDS and used for western blot analysis.

Whole organelle immunoprecipitation

Intact Golgi and E.R. fractions were harvested from approximately 10⁸ HeLa cells using discontinuous sucrose gradients as described above. Organelles were then immunoprecipitated using Dynabead Protein G magnetic beads (Invitrogen, 100.03D). Beads were first washed twice with citrate-phosphate buffer (25 mM Citric Acid, 52 mM Na₂HPO_4, pH 5.0) using a magnet to pellet. Approximately 5 ug of antibody was added to each organelle fraction followed by 50 ul of the magnetic beads. The mixture was then allowed to rotate at 4°C for 1 h. The beads were then gently washed 3 times with PBS using a magnet to pellet. After the final wash, the beads were boiled in 2X SDS to be used for western blot.

CDC42 Activation Assay

The level of endogenous GTP-bound CDC42 was determined by affinity precipitation assay as previously described³⁸ using the Cdc42 binding domain of PAK3.

Golgi disruption assay

HeLa cells were grown on glass coverslips and transfected with indicated constructs. Twenty-four hours post transfection, cells were treated with Brefeldin A (10 ug/mL, Invitrogen, B7450) for 30 min at 37°C. Cells were then stained with a Golgi-specific marker (Molecular Probes, W-11261), DAPI (Calbiochem, 268298), and antibodies that recognize the relevant overexpressed protein. Cells expressing the appropriate markers were then scored (>50) for the presence or absence of an intact Golgi.

VSVG assay

HeLa cells were transfected with VSVG-GFP along with the Dbs-Sec14 construct or cognate vector. Twenty-four hours later cells were cultured at 40°C for 12 h to trap the VSVG protein in the ER. Cultures were then shifted to 32°C incubator for 20 min or 90 min to allow transport of the VSVG protein to the Golgi and plasma membrane respectively. Cells were then fixed and stained for expression of the inhibitor. Quantification of VSVG transport was performed by manually counting the number of cells expressing VSVG which have plasma membrane, or Golgi, localization out of the total number of cells expressing VSVG. Counts were performed blinded by 2 independent investigators.

Transwell motility assay

The transwell motility assay was performed as described previously.³⁹ Briefly, subconfluent MDA-MB-231 cells were transfected with either the pEGFP-Sec14 construct or cognate vector. After 24 h, GFP positive cells were sorted by FACS analysis and 50 000 GFP positive cells were loaded onto transwells coated from the underside with Type I Rat Tail Collagen (BD Biosciences, 354236). After 19 h, cells that had passed through the filter were fixed and stained with the Diff-Quik Stain Set (Siemens, B4132–1A).

In vitro wound-healing assay

Wound-healing assays were performed as described by others.⁴⁰ MDA-MB-231 cells were grown on 35mm tissue culture dishes coated with 50 ug/mL Type I Rat Tail Collagen (BD Biosciences, 354236). Cells were transfected with pEGFP-Sec14, or cognate vector, and allowed to grow to confluence. Cells were serum-starved for 4 h, and then the monolayer was wounded using a sterile pipette tip. Complete media was then added back to the cells and migration was allowed to take place for 19 h. Closure of the wound was monitored and imaged at >10 locations for each condition. GFP positive cells in the vicinity of the wound were counted before migration. After 19 h, GFP positive cells found in the wound closure were scored as positive for migration.

Golgi orientation assay

HeLa or MDA-MB-231 cells were grown on glass coverslips, transfected with pEGFP-DBS-Sec14, or cognate vector, and then allowed to grow to confluence. The monolayer was wounded using a pipette tip and then cells were allowed to orient for 3 h. Cells were then fixed and stained with a Golgi-specific marker, an actin-specific marker (phalloidin), and antibodies that recognize the GFP-tag. Cells were then imaged along the wound edge. A cell that had the majority of its Golgi located in the 120 degree sector facing the wound was considered positive for orientation.

Statistics

P values were determined using a Student's T test for non-paired values. Values are typically given as means ± SE.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

This work was supported by Public Health Services Grant CA97066 from the NCI, National Institutes of Health, and a Research Scholar Grant RGS-04–199–01 from the American Cancer Society (I.P.W.). E.F. was supported by a predoctoral fellowship from the New Jersey Commission on Cancer Research.