H-Ras Does Not Need COP I- or COP II-dependent Vesicular Transport to Reach the Plasma Membrane*

Although vesicular transport of the H-Ras protein from the Golgi to the plasma membrane is well known, additional trafficking steps, both to and from the plasma membrane, have also been described. Notably, both vesicular and nonvesicular transport mechanisms have been proposed. The initial trafficking of H-Ras to the plasma membrane was therefore examined in more detail. In untreated cells, H-Ras appeared at the plasma membrane more rapidly than a protein carried by the conventional exocytic pathway, and no H-Ras was visible on Golgi membranes in >80% of the cells. H-Ras was still able to reach the plasma membrane when COP II-directed transport was disrupted by two different mutant forms of Sar1, when COP I-mediated vesicular traffic from the endoplasmic reticulum to the Golgi was inhibited with brefeldin A, or when microtubules were disrupted by nocodazole. Although some H-Ras was present in the secretory pathway, protein that reached the membranes of the endoplasmic reticulum-Golgi intermediate compartment was unable to move further in the presence of nocodozale. These results identify an alternative mechanism for H-Ras trafficking that circumvents conventional COPI-, COPII-, and microtubule-dependent vesicular transport. Thus, H-Ras has two simultaneous but distinct means of transport and need not depend on vesicular trafficking for its delivery to the plasma membrane.

The broad outlines of the biosynthetic transport of Ras proteins to the plasma membrane are known (1)(2)(3). Mammalian H-Ras, N-Ras, and K-Ras and also yeast Ras proteins are initially cytosolic, become modified with a farnesyl isoprenoid, and are then escorted, perhaps by the farnesyltransferase itself (4,5), to the cytosolic face of the widely dispersed membranes of the endoplasmic reticulum. There, farnesylated Ras proteins are further modified by an ER 2 -resident prenyl-dependent protease (Rce1), and by a second ER-resident methyltransferase enzyme (Icmt) (6,7). At this point the K-Ras4B protein is released from the surface of the ER and rapidly appears at the plasma membrane (3,8,9). No internal membrane or vesicular intermediates have been observed for this step, and the transport mechanism is currently unknown (10). In neural cells, calcium signals cause calmodulin to bind to plasma membrane-bound K-Ras4B and trigger its internalization through a protein-mediated, cytoplasmic shuttle mechanism (11). Thus K-Ras4B, at least in neurons, appears to use nonvesicular means for both initial outward and signal-dependent inward trafficking.
For H-Ras, N-Ras, and yeast Ras2, a second lipid modification can occur, in which a palmitoyl lipid is attached to additional cysteines near the farnesylated C terminus. A two-subunit protein acyltransferase enzyme that attaches palmitates to H-Ras and N-Ras in mammalian cells has only recently been identified and may be located on ER and Golgi membranes (12)(13)(14). In yeast, a similar two-subunit enzyme (Erf2/Erf4) that modifies yeast Ras2 has been found on ER membranes (15). Another mammalian acyltransferase, Huntington interacting protein-14 , is distributed on ER/Golgi membranes and has been reported to enhance vesicular trafficking of multiple acylated proteins (16,17). Additional palmitoyltransferase enzymes are likely to be located in the plasma membrane, because mutant yeast and mammalian Ras proteins that reach the plasma membrane without associating with ER or Golgi membranes can be palmitoylated (18,19).
The movement of H-Ras from the cytosolic surface of the vast ER membrane network is the first step in its journey to the plasma membrane. Two recent reports have laid the framework of this path. These studies propose that nonpalmitoylated (but farnesylated) H-Ras distributes rapidly between cytosolic and transient, nonspecific membrane-bound phases. Wherever H-Ras encounters an acyltransferase and is palmitoylated, it becomes "kinetically trapped" on the membranes (20) and, as a result, can then be retained on vesicles and moved onward via traditional vesicular means (21,22). Proteins that control vesicular trafficking at the sites where acyltransferases are found thus become relevant for H-Ras transport. However, the model currently does not address whether a farnesylated H-Ras will need a partner protein to mask its lipid group during its cytosolic phase. These possibilities give importance to understanding the steps that H-Ras takes as it makes its way from the ER to the plasma membrane.
A great deal is known about vesicular trafficking from the ER to the Golgi. Formation of vesicles on the ER begins with assembly of a COP II complex of proteins around the site, termed an ER exit site (ERES), at which the newly cargo-loaded vesicle will form (23)(24)(25). The protein that initiates this event is the GTPase, Sar1. GTP-binding mutants of Sar1 disrupt this cycle, * This work was supported by National Science Foundation Award 0110114.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains a supplemental movie. 1 To whom correspondence should be addressed. and inhibit COP II-mediated vesicle transport from the ER to Golgi (26). In mammalian cells, the COP II vesicles that are released from the widely dispersed ERES then seem to merge into a perinuclear vesiculo-tubular center sometimes termed the ER-Golgi-intermediate compartment (ERGIC). From here, some vesicles move onward to the Golgi, whereas others return to the ER (27). COP I-coated vesicles mediate anterograde transport from the ERGIC to Golgi and retrograde traffic from Golgi membranes to the ER (28,29). The fungal product brefeldin A (BFA) prevents the COP I complex from binding to membranes. This causes collapse of the Golgi stacks, as the membranes and proteins of the Golgi rapidly drain backwards into the ER, and also impairs anterograde trafficking beyond the ERGIC (30). Thus, BFA inhibits vesicular transport at a separate point and by a mechanism distinct from that of Sar1 mutants.
BFA has been reported to partially inhibit transport of a chimeric GFP protein with a lipidated C terminus derived from H-Ras (3). However, another study reported continued transport of both H-Ras and another GTPase, TC10, to the plasma membrane in BFA-treated cells (31). Thus, it is unresolved to what extent BFA-sensitive, conventional vesicular trafficking contributes to H-Ras transport. The yeast Ras2 protein (which is modified by both farnesyl isoprenoid and palmitoyl lipids) has been reported to use a nonconventional pathway for trafficking (19).
These precedents prompted us to examine the mechanisms by which mammalian H-Ras accesses the plasma membrane, with particular attention to the initial steps at its release from the ER. The results indicate that two distinct routes for H-Ras transport operate simultaneously, and that a nonconventional, COP I-and COP II-independent mechanism moves the bulk of H-Ras to the plasma membrane. Unexpectedly, when traditional vesicular trafficking is blocked, H-Ras that has entered that pathway appears unable to switch to the nonconventional pathway. This suggests that the choice of pathway, and the mechanism of transport, is determined prior to ER disengagement.

EXPERIMENTAL PROCEDURES
Mutants, Plasmids, and Antibodies-H-Ras wt , H-Ras Q61L , and GFP-H-Ras were expressed from pcDNA3 (Invitrogen). The cDNAs for YFP-GT46 and the monomeric GFP (mGFP) (32) were gifts from Anne Kenworthy (Vanderbilt University), with permission of Roger Tsien (University of California, San Diego). The cDNAs for hemagglutinin (HA)-tagged Sar1 T39N -HA and Sar1 H79G -HA were gifts from Dr. Jennifer Lippincott-Schwartz (National Institutes of Health). Anti-Ras rat monoclonal antibody 238 (Santa Cruz Biotechnology), anti-HA mouse monoclonal antibody (Covance), anti-Giantin rabbit polyclonal antibody (Covance), anti-ERGIC-53 mouse monoclonal antibody (Alexis), and AlexaFluor-conjugated secondary antibodies (Molecular Probe) were used for immunofluorescence. The anti-Ras rat monoclonal antibody was detected with a rat-specific secondary antibody. YFP-GT46 is composed of a signal sequence, enhanced yellow fluorescent protein as an ectodomain, a sequence to direct N-glycosylation, the transmembrane domain of the human low density lipoprotein receptor, and the cytoplasmic tail of CD46 (33).
Cell Culture and Transfection-NIH 3T3 cells and COS-7 cells were cultured at 10% CO 2 in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% calf serum (Hyclone Laboratories). Twenty four hours before DNA transfection, cells were plated on pre-washed glass coverslips (Corning) in 12-well tissue culture plates at a density about 10 5 -10 6 cells per well. Transfection was performed using Effectene reagent (Invitrogen) as described by the manufacturer. For sequential DNA transfection, only Sar1 DNA was used for the first transfection, and after 18 h a second round of transfection, with DNA for H-Ras, was performed. For Sar1 co-transfection, a DNA ratio of 2:1 (Sar1/other DNAs; w/w) was used. For all other co-transfections a 1:1 DNA ratio (w/w) was used.
Counting Cells-NIH 3T3 cells were co-transfected with cDNAs for YFP-GT46 and H-Ras, as indicated. After 18 h, cells were fixed and viewed by epifluorescence (for GFP constructs) or immunostained with antibody for H-Ras or Giantin. One hundred cells with YFP-GT46 (or Giantin) visible in the perinuclear area were then counted, and the number of cells in which H-Ras was present only on plasma membrane or coincident with the perinuclear marker protein was noted. Two separate experiments were performed for each H-Ras form, and the data were analyzed using the GraphPad Prism software program.
Brefeldin A, Nocodazole, and Cycloheximide Treatment-At 6 h after transfection, brefeldin A (BFA, 5 g/ml) was applied to the cells in regular medium. Cells were fixed and visualized by immunofluorescence at 30, 60, and 90 min after BFA treatment. Where indicated, cells were treated with cycloheximide (50 g/ml) immediately after the transfection. After 6 h the cycloheximide was washed out, and fresh medium with either BFA or nocodazole (20 g/ml) was added. After 1.5 h, cells were fixed for immunofluorescence. For one experiment, cells were transfected, treated with cycloheximide for 6 h, and then were rinsed to wash out the cycloheximide, and nocodazole was applied for 4 h. Then nocodazole was washed out, and cycloheximide was added to the medium again. After 2 more hours, cells were fixed and stained for immunofluorescence.
Immunofluorescence Imaging-NIH 3T3 and COS-7 cells were fixed with 4% formaldehyde in phosphate-buffered saline (PBS) at room temperature for 15 min. Cells were permeabilized with 0.2% Triton X-100 on ice for 5 min and treated with 50 mM NH 4 Cl in PBS for 15 min at room temperature. Nonspecific binding was blocked using 1% fat-free milk in PBS for 15 min, followed by primary antibody for 1 h and secondary antibody for 1 h at room temperature. Coverslips were mounted with VECTASHIELD, and images were captured with a Leica inverted fluorescence microscope (model DIMER2) or Leica confocal TCS NT microscope system equipped with separate argon-UV, argon, and krypton lasers and the appropriate filter sets for fluorescein isothiocyanate, Texas Red, and TRITC imaging. The contrast and signal strength of images were balanced, and images were deconvolved and merged as indicated, using Improvision OpenLab and Adobe Photoshop software.
Supplemental Movie-NIH 3T3 cells were cultured and transfected with GFP-H-Ras cDNA as described before. Cells were treated with cycloheximide (50 g/ml) immediately after the transfection. After 6 h the cycloheximide was washed out, and fresh medium with BFA (5 g/ml) and HEPES (20 g/ml, pH 7.4) was added. Cells were thereafter kept at 37°C in a warmed chamber. After 30 min of BFA treatment, images of a selected cell were taken at 90-s intervals for about 1 h with a Leica inverted fluorescence microscope (model DIMER2). Movies were created by Improvision OpenLab software.

H-Ras Proteins Often Show Little Accumulation in the Golgi-
Although previous reports clearly documented the presence of H-Ras on the ER and Golgi, most of those studies used GFP-H-Ras fusion proteins to produce bright signals when localized to a small area (1,3). When a nontagged version of H-Ras was imaged, using an H-Ras antibody, many transfected NIH 3T3 cells showed H-Ras protein only on the plasma membrane, with no protein detectable on perinuclear membranes. This visual impression was quantified by determining the proportion of cells that showed H-Ras immunofluorescence on Golgi membranes. To verify the location of the Golgi apparatus, a co-expressed YFP-GT46 protein, which traffics through the conventional exocytic pathway (34), or Giantin, a resident cis-Golgi protein, were also imaged. At 18 h after transfection, H-Ras was strongly visible on the plasma membrane in Ͼ95% of all transfected cells. YFP-GT46 was detectable at the plasma membrane, and also in an organized perinuclear area in Ͼ90% of the transfected cells (Table 1). One hundred cells that showed YFP-GT46 in the perinuclear area were then counted, and the presence and abundance of H-Ras on membranes that were coincident with the perinuclear YFP-GT46 were noted in each. The results verified the preliminary impression that H-Ras was frequently absent from internal membranes (Golgi or recycling center (35)), as ϳ85% of the transfected cells showed no H-Ras on organized internal membrane structures that were also marked by YFP-GT46. Roughly 10% of the cells showed some modest amount of internal H-Ras, whereas about 5% showed strong co-localization with perinuclear membranes containing YFP-GT46. The same experiment was performed using COS-7 cells, with similar results (Table 1). Thus, in both cell types, only a few of the transfected cells showed a strong H-Ras signal on internal membranes.
To see if a GTP-bound, oncogenic form of H-Ras behaved similarly, an activated version, H-Ras Q61L was expressed. Once more, in 85-95% of the transfected cells, there was very little H-Ras Q61L that co-localized with YFP-GT46 in the perinuclear area, although the plasma membrane was illuminated ( Fig. 1A; Table 1). Thus, both the native form and an oncogenic form of H-Ras displayed limited accumulation on internal membranes.
To look more specifically at the Golgi, the experiment was repeated, using COS-7 cells and staining of the endogenous protein, Giantin, to mark the cis-Golgi. Again, a large majority of the cells showed almost no H-Ras Q61L on Golgi membranes ( Fig. 1B; Table 1). A similar limitation in Golgi localization of H-Ras has been reported in CHO-K1 cells.
In contrast, when GFP-H-Ras fusion proteins were examined, a larger portion (ϳ70%) of transfected cells showed the GFP-H-Ras fusion proteins on Golgi membranes, and only ϳ30% of the cells lacked any accumulation of H-Ras in the Golgi area (Table 1). This significant amount of GFP-H-Ras on Golgi membranes agrees with previous work using GFP-H-Ras proteins (1). To learn if this apparent influence of GFP on H-Ras localization might arise because of the tendency for the original versions of GFP to dimerize, a monomeric form of GFP was attached to H-Ras Q61L , and the steady-state distribution of this mGFP-H-Ras Q61L was examined. In roughly half of the cells, mGFP-H-Ras Q61L was visible on perinuclear membranes, compared with ϳ65% for the original GFP version. Meanwhile, the portion of cells in which no perinuclear fluorescence of mGFP-H-Ras was visible increased moderately, from ϳ35% for the original GFP-tagged version to about 50% with the monomeric GFP version (Table 1). However, the mGFP-H-Ras Q61L was still clearly different in its distribution from that of the untagged H-Ras Q61L , in which Ͻ5% of the cells showed perinuclear fluorescence and Ͼ90% did not. These results suggest that, although GFP-H-Ras (or mGFP-H-Ras) proteins may be useful for documenting the presence of H-Ras at internal locations, they may not fully replicate normal usage of the transport pathways.

Nascent H-Ras Moves Much
Faster than YFP-GT46 to the Plasma Membrane-These results suggested two possibilities as follows: (a) in 85% of the cells the amount of H-Ras on Golgi membranes was low because vesicular transport was particularly efficient; or (b) a significant amount of H-Ras might be able to access the plasma membrane without utilizing the exocytic pathway through the Golgi.
To study trafficking efficiency, the initial movement of H-Ras was examined at early time points after the protein was synthesized. This avoided the complexity of possible deacylationbased recycling of previously synthesized H-Ras between plasma and internal membranes (21). For this, cells were cotransfected with cDNAs for H-Ras Q61L and YFP-GT46 and then were placed immediately in medium containing cycloheximide for 6 h to allow the cDNAs to be transcribed and the RNA for H-Ras to accumulate, but to prevent translation. Cycloheximide was then removed to produce a synchronous wave of H-Ras protein synthesis and trafficking. At various time points after cycloheximide removal, cells were fixed and imaged. Using this approach, both H-Ras and YFP-GT46 had been translated and accumulated in amounts sufficient to be visible within 50 min of removal of cycloheximide. At that time, both proteins were present only on the ER (Fig. 2, 50Ј). Twenty minutes later (70 min after cycloheximide removal), a strong signal of YFP-GT46 was present on well organized perinuclear membranes in ϳ90% of the transfected cells (Fig. 2, GT46, 70Ј). Importantly, no visible YFP-GT46 had yet reached the plasma membrane. In contrast, H-Ras was already detectable on the cell surface, well ahead of YFP-GT46 (Fig. 2, H-Ras, 70Ј). By 90 min after cycloheximide removal, YFP-GT46 was present in the Golgi and, in some cells, on vesicles that appeared to be exiting the Golgi but were not yet at the cell surface (Fig. 2, GT46, 90Ј). At this time point, H-Ras was now easily visible at the plasma membrane (Fig. 2, H-Ras, 90Ј), although no H-Ras-containing vesicles were detected and fewer than half of the cells showed any H-Ras in the perinuclear area. After another 20 min (110 min after cycloheximide removal), exocytic vesicles containing YFP-GT46 could be seen in most transfected cells. In some cells, these GT46containing vesicles had reached the cell surface, and the outline of these cells could now be barely seen (Fig. 2, GT46, 110Ј). H-Ras was now present on the plasma membrane and on both perinuclear membranes and individual "granules" inside the cell (Fig. 2, H-Ras, 110Ј). When the images were merged, many of the vesicles containing YFP-GT46 did not have H-Ras on them (Fig. 2, merge, 110Ј). This result suggested that much of the H-Ras was not using YFP-GT46containing vesicles for transportation to the plasma membrane. By 140 min after cycloheximide removal, large amounts of YFP-GT46 had reached the plasma membrane, and many YFP-GT46-containing vesicles en route were obvious. Merged images of H-Ras and GT46 showed some vesicles that contained both H-Ras and YFP-GT46, although again, many contained only one of the proteins (Fig. 2, merge, 140Ј). The co-localization of YFP-GT46 and H-Ras agrees with previous reports that (some) H-Ras can traffic from the Golgi to the plasma membrane on vesicles (1,21,22).
These results clearly showed that nascent H-Ras could reach the plasma membrane more rapidly than a protein carried on the traditional vesicular pathway, Furthermore, in many cells, there was no evidence that H-Ras transited internal membranes. This led us to test whether H-Ras could reach the plasma membrane if conventional vesicular transport was prevented.

Neither Brefeldin A nor Dominant Negative Sar1 Prevents H-Ras Movement to the Plasma Membrane-To test if H-Ras
plasma membrane traffic required the presence or organization of the Golgi apparatus, brefeldin A was used to inhibit COP I-mediated vesicular transport and cause Golgi collapse. Cells were transfected with cDNAs for H-Ras Q61L and YFP-GT46 and were immediately treated with cycloheximide. After allowing 6 h for RNA synthesis, the cycloheximide was washed out to permit protein synthesis, but BFA was also added. After an additional 90 min, the cells were fixed and imaged. As expected for a transmembrane protein in the secretory pathway, YFP-GT46 was found in the ER after BFA treatment and clearly outlined the ER membranes around the nucleus (Fig. 3A). No other internal membranes contained YFP-GT46, confirming that vesicular traffic had halted, and the Golgi apparatus had been absorbed into the ER. In contrast, H-Ras Q61L continued to illuminate the plasma membrane in the presence of BFA (Fig.  3A). No organized perinuclear structures that contained  AUGUST 31, 2007 • VOLUME 282 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 25763
H-Ras Q61L were visible, and the protein did not accumulate on the nuclear envelope. To determine whether the activation state of the protein influenced whether H-Ras transport was BFA-insensitive, the experiments were repeated using the inactive, cellular form of H-Ras (H-Ras wt ). The results were similar to those with the activated form. BFA had no inhibitory effect on movement of H-Ras wt to the plasma membrane (data not shown).
Additional trials with a 3-h exposure to BFA were performed to examine if H-Ras might begin to accumulate on ER membranes if given more time. Even with this longer treatment, both H-Ras wt and H-Ras Q61L continued to move to the plasma membrane (data not shown). Also, BFA added to cells for 4 h at 18 h after transfection caused no accumulation of H-Ras on internal membranes (data not shown). Additional experiments with H-Ras Q61L expressed in COS-7 cells showed that H-Ras Q61L movement to the plasma membrane was not prevented by BFA in this other cell type (data not shown). These results indicated that neither GDP-bound nor GTP-bound forms of H-Ras required physically distinct Golgi membranes or COP I-coated vesicles in order to travel to the cell surface.
Although the results with BFA suggested that H-Ras might not require vesicular transport to reach the plasma membrane, these studies inhibited only the COP I side of this process. The necessity for H-Ras vesicular trafficking was tested in a second way, by preventing assembly of the COP II complex, using mutants of the assembly regulator, Sar1. Cells were co-transfected with cDNAs for H-Ras Q61L and a GDP-bound inhibitory form of Sar1 (Sar1 T39N -HA), and 24 h later were fixed and imaged. The H-Ras Q61L protein strongly outlined the plasma membrane of cells that overexpressed the Sar1 dominant negative protein (Fig. 3B). There was no indication of accumulation of H-Ras Q61L in the ER or around the nuclear envelope, nor any Golgi-like structures. This suggested indirectly that the Sar1 T39N had successfully ablated the Golgi. A second, longer experiment was also performed to give the Sar1 T39N more time to amass to levels that would certainly disrupt ERES before H-Ras was present. Even after 48 h, in cells co-expressing Sar1 T39N and H-Ras Q61L , the H-Ras Q61L continued to mark the plasma membrane and showed no build up in the ER (data not shown). In addition, a sequential transfection technique was developed to allow Sar1 more time to become functional before H-Ras was synthesized and began its trafficking. Cells were first transfected with cDNA for the Sar1 T39N and allowed to express this protein for 18 h. The cells were then transfected a second time with cDNA for H-Ras Q61L and after a further 24 h were fixed and imaged. All cells that expressed H-Ras Q61L showed strong plasma membrane staining, including those that also co-expressed the previously synthesized Sar1 T39N (Fig.  3B). In cells that expressed both Sar1 T39N and H-Ras Q61L proteins, there was no Golgi-like staining of H-Ras Q61L on internal membranes. In cells that expressed only H-Ras Q61L or that had only low expression of Sar1 T39N , there was again a mixture of cells displaying H-Ras on plasma membrane only or having H-Ras on both plasma membrane and perinuclear membranes. H-Ras Q61L did illuminate some scattered internal membranes that did not have the layered appearance of Golgi and seemed to be fluid-filled pinosomes, with H-Rasstained rims but dark interiors (Fig. 3B). Activation of H-Ras is known to strongly stimulate fluid-phase endocytosis (37), so these structures may be the recently reported H-Ras Q61Lcontaining endocytic vesicles that had been internalized after reaching the plasma membrane.
To ensure that the Sar1 T39N protein was functioning correctly, a similar experiment was performed using the YFP-GT46 protein. The mutant Sar1 prevented YFP-GT46 from trafficking to the plasma membrane and caused it to accumulate on internal ER and nuclear envelope membranes (Fig. 3B, lower row). This indicated that the amount of Sar1 T39N that was expressed was sufficient to block transport of a protein in the traditional exocytic pathway. Equivalent results were obtained with Sar1 T39N and H-Ras Q61L expression in COS-7 cells (data not shown), indicating that H-Ras trafficking is Sar1-independent in other cell types.
To disrupt exit site assembly at a separate step, a different GTP-bound Sar1 mutant, Sar1 H79G -HA, was employed. With either simultaneous or sequential transfection protocols, H-Ras Q61L continued to reach the plasma membrane in cells that also expressed the Sar1 H79G -HA protein (data not shown). These results indicated that H-Ras did not need COP II vesicles in order to traffic to the plasma membrane.

Structures That Carry H-Ras Do Not Show Directed Movement to the Plasma Membrane in BFA-treated Cells-
The results to this point suggested that a great deal of H-Ras trafficking occurred via a pathway to the plasma membrane that was independent of the classical vesicular route through the Golgi. To examine this alternative pathway in more detail, the classical pathway was ablated with BFA, and a movie was recorded to study if any type of BFA-resistant vesicles carrying H-Ras protein could be seen moving to the plasma membrane. NIH 3T3 cells were transfected with GFP-H-Ras Q61L cDNA and treated with cycloheximide immediately after the transfection. After 6 h, cycloheximide was washed out, and BFA was added to the medium. After another 30 min to allow protein synthesis, the culture dish was moved to a chamber held at 37°C, and movement of the newly synthesized GFP-H-Ras Q61L was recorded. As had been seen in the earlier experiments with fixed cells, at 40 min after the wash out of cycloheximide, detectable amounts of GFP-H-Ras Q61L had already appeared in the cytosol and on ER membranes. The GFP-H-Ras Q61L was also present on numerous, very small, granular structures within the cell. These granules were unlikely to be conventional vesicles as they did not arise from any organized perinuclear, Golgi-like center, and the previous experiments had shown that this time point was prior to the time at which (limited amounts of) H-Ras would be seen on conventional exocytic vesicles. The GFP-H-Ras Q61L granules moved slowly, often inward (Fig. 4,  insets), and had no specific direction toward the plasma membrane ( Fig. 4 and supplemental movie). A few granules (two in the eight cells that were imaged) did fuse with the plasma membrane and illuminated the membrane temporarily (data not shown). Notably, the GFP-tagged H-Ras protein could be seen on ruffles and plasma membrane of the BFA-treated cells. Thus, a GFP-tagged version of H-Ras could successfully access the plasma membrane during BFA exposure. This indicated that the GFP tag did not prevent this type of trafficking. As more GFP-H-Ras protein was synthesized after the removal of the cycloheximide, an increase in GFP-H-Ras fluorescence was noted in the perinuclear area. This area was not the Golgi, because it had been ablated by the BFA that was present. The identity of this region was examined below. The general absence of directed movement of the visible dots of GFP-H-Ras Q61L , in addition to its COP I and COP II independence, suggested that the initial transport of GFP-H-Ras Q61L to the plasma membrane that occurred in BFA-treated cells might be nonvesicular.

Nocodazole Does Not Prevent H-Ras Movement to the Plasma
Membrane-To determine whether transport of nascent H-Ras was also independent of microtubules that might support vesicular traffic, cells transfected with cDNA for H-Ras Q61L were treated with cycloheximide for 6 h; cycloheximide was removed, and nocodazole was added for 90 min. With this treatment length, nocodazole causes fragmentation of the Golgi into multiple individual Golgi mini-stacks throughout the cell. These stacks were visualized with antibody to the endogenous Golgi protein, Giantin, which verified that the nocodazole treatment had successfully dispersed the Golgi (Fig. 5). To determine how effectively this nocodazole treatment inhibited transport of a protein in the traditional exocytic pathway, GT46 was examined. After a 4-h exposure to nocodazole, GT46 was found exclusively on internal membranes (Fig. 5); no GT46 was visible on the plasma membrane. This indicated that this protocol for nocodazole treatment efficiently inhibited microtubule-dependent vesicular transport during the experiment. Importantly, in cells in which microtubules had been depolymerized, H-Ras Q61L was still able to reach the plasma membrane (Fig. 5). This indicated that a microtubule-independent mechanism could move H-Ras to the plasma membrane. These results complement previous work (22) that had shown that H-Ras that was already at the plasma membrane could reach the Golgi, when microtubules were depolymerized by acute application of nocodazole.
Notably, in these nocodazole-treated cells, there was a prominent amount of H-Ras that was located on dispersed internal structures. These HRas-containing structures were often alongside of, but did not actually co-localize with, the Giantinmarked Golgi mini-stacks (Fig. 5, merge). This revealed that, although a significant amount of H-Ras reached the plasma membrane, nocodazole treatment caused some H-Ras to be retained internally, on structures that were distinct from cis-Golgi membranes in the trafficking pathway. These results pointed to the possibility of simultaneous utilization of two parallel mechanisms for H-Ras transport. One route was the traditional vesicular path, dependent upon microtubule integrity; the other was a pathway that did not require COP I or COP II vesicles or microtubules. Some H-Ras Accumulates at the ERGIC When the Classical Pathway Is Blocked-The identity of the non-Golgi structures that contained H-Ras in nocodazole-treated cells was examined. A clear candidate location was the series of membranes, often termed the ERGIC (38,39), that bud off the widely dispersed ER exit sites and move toward the Golgi. Microtubule de-polymerization does not prevent traffic from ER to ERGIC, but it does interrupt vesicular transport from ERGIC to Golgi (39). Thus, at steady state the bulk of ERGIC membranes are located near the perinuclear cis-Golgi, but in nocodazoletreated cells, proteins in the ERGIC reside near the ERES and Golgi mini-stacks. ERGIC membranes are also BFA-resistant, are maintained rather than absorbed during BFA exposure, and even accumulate in a perinuclear mass in cells treated for extended periods with BFA (39). Thus, both nocodazole and BFA cause proteins in the conventional trafficking pathway to accumulate in the ERGIC, although the ERGIC membranes are organized in spatially distinct ways by these drugs.
To learn if some H-Ras accumulated on ERGIC membranes in cells in which vesicular transport was impaired, transfected COS-7 cells were treated with cycloheximide for 6 h, then with nocodazole or BFA for 1.5 h in the absence of cycloheximide, and finally were fixed and stained. Antibody to the endogenous ERGIC-53 protein was used to mark the location of ERGIC membranes. In untreated cells, the ERGIC-53 protein was located on a cluster of perinuclear membranes (Fig. 6, control). The ERGIC-53-containing membranes intertwined around the Giantin-marked cis-Golgi (data not shown), making these structures difficult to resolve in untreated cells. In the minority of untreated cells that showed H-Ras on perinuclear membranes, the newly synthesized H-Ras Q61L protein showed substantial co-localization with the endogenous ERGIC-53 (Fig. 6,  control).
In agreement with previous reports, in nocodazole-treated cells, the ERGIC-53 antibody stained membranes that had dispersed throughout the cell body (Fig. 6, ϩNoc), although in BFA-treated cells, ERGIC-53 staining appeared on a cluster of perinuclear membranes (Fig. 6, ϩBFA). In cells treated with either nocodazole or BFA, the internal sites where the H-Ras Q61L protein was located co-localized with the ERGIC-53 protein (Fig. 6). Once again, H-Ras was also present on the plasma membrane in these nocodazole-or BFA-treated cells. There was also a considerable amount of the newly synthesized H-Ras that was still cytosolic or on the widely dispersed membranes of the ER. These results indicated that some of the newly synthesized H-Ras had entered the traditional trafficking pathway and accumulated in the ERGIC when vesicular transport was impaired.

H-Ras That Is in the ERGIC Does Not Access the Alternate Pathway in Nocodazole-treated Cells-If
H-Ras could move to the plasma membrane via a nonconventional mechanism that originated prior to the Golgi, it might enter this alternate pathway directly from the ER, shortly after interacting with the enzymes that remodel the C terminus. Alternatively, it might aggregate at ER exit sites and move to the ERGIC via conventional means, but from there disengage and be routed to the cell surface. The proteins that might control H-Ras transport would likely differ at these two locations. The failure of inhibitory Sar1 proteins to prevent H-Ras plasma membrane access suggested that the alternate pathway did not require the proteins found at ER exit sites. Another experiment was designed to test if the H-Ras that accumulated in the ERGIC upon nocodazole treatment could eventually access the plasma membrane. COS-7 cells were transfected, and cycloheximide was added immediately. After 6 h, cycloheximide was washed out to generate synchronous production of nascent H-Ras proteins, and nocodazole was added at that same time to allow some of these nascent H-Ras proteins to get into, but to impair their exit from, the ERGIC. After 4 h, when a detectable portion of H-Ras protein accumulated in the ERGIC, a second cycloheximide treatment was applied, to prevent synthesis of additional H-Ras, and the cells, still in the presence of nocodazole, were incubated a further 2 h. Thus, only pre-existing H-Ras protein was in the ERGIC, and new protein could not move in. The prediction was that if H-Ras could be exported from ERGIC via the alternative pathway, it would disappear from the ERGIC because there would be no new H-Ras supplied from the ER. However, this was not observed. H-Ras was still present in the membranes that aligned with ERGIC-53 even after 2 more hours (Fig. 7, ϩNoc Chase). In a control experiment, nocodazole was removed during the second cycloheximide exposure, to now allow vesicular traffic out of the ERGIC. During this nocodazole wash out, the organized, perinuclear structure of the ERGIC quickly re-formed, and H-Ras was observed on many internal vesicle-like structures that no longer co-aligned with ERGIC-53 (Fig. 7, ϪNoc Chase). This confirmed that the H-Ras that had become trapped during the nocodazole treatment had not been damaged and remained competent for transport. Thus H-Ras could rapidly leave the ERGIC if vesicular transport was available, but it could leave only very slowly or not at all if that conventional mode of transport was impaired. This result suggests that the two modes of transport are separate and that the H-Ras that reaches the ERGIC can no longer access the alternate route. This supports the possibility that this alternate pathway for H-Ras transport to the plasma membrane originates directly from the ER.

DISCUSSION
The results here represent the first comprehensive look at modes of transport used to move newly synthesized H-Ras to the plasma membrane. Despite our expectations, conventional vesicular mechanisms could not account for most of the H-Ras that reached the cell surface. Previous work, although focused on the traditional exocytic pathway, had also noted that a significant amount of H-Ras reached the plasma membrane when traditional trafficking pathways were inhibited (3). Fluorescence photo-bleaching experiments had also indicated that the H-Ras protein moves from the ER to the Golgi too rapidly to be carried by vesicles (22).
Those studies, combined with the evidence presented here, lead us to propose that a significant amount of H-Ras uses a nonclassical pathway for transportation to the plasma membrane. This model proposes that H-Ras has two pathways by which it can reach the plasma membrane, and that in normal circumstances both routes are used simultaneously. In NIH 3T3 or COS-7 cells, this nonclassical pathway appears to be the primary means for plasma membrane access. In yeast, strong genetic evidence supports the idea that the Ras2 protein moves to the cell surface via a nonconventional pathway (19,40). This indicates that a nonclassical pathway of Ras transport may be conserved in eukaryotes.
This nonclassical pathway appears to be initiated at the ER rather than at the Golgi. Comparison of the localizations of H-Ras and Sec13 on ER membranes suggests that very little H-Ras is present at conventional vesicle budding domains of ER export sites. 3 In addition, once H-Ras enters the conventional transport pathway and reaches the ERGIC, it appears to have very limited access to this nonclassical pathway. This suggests that the choice of pathway, and mechanism of transport, is determined prior to ER disengagement. H-Ras thus has two distinct means of transport and, importantly, does not need to depend on vesicular trafficking for its delivery to the plasma membrane.
Although the two different pathways for H-Ras traffic appear to separate at the ER, it is not yet known what initiates this divergence. One strong candidate for such a mechanism is palmitoylation (40). There is existing evidence that, as long as it is nonpalmitoylated, farnesylated H-Ras is in a dynamic equilibrium between the cytosol and membranes (20,22,41). Current models propose that when H-Ras encounters a membranebound palmitoyltransferase and becomes palmitoylated, it can no longer disengage from that membrane. The location at which palmitoylation occurs then defines the point where the palmitoylated H-Ras can join the classical vesicular pathway (21,22). Our evidence that some H-Ras is in the secretory pathway and can reach ERGIC membranes, but in nocodazoletreated cells appears unable to leave, provides independent support for this model. The data also bolster the recent report that a palmitoyltransferase may be located on ER membranes (13) and also suggest that it may be located on perinuclear ERGIC membranes as well. Our model would propose that regulation of the activity of this enzyme might then adjust the amounts of H-Ras that enter each of the two transport pathways. If H-Ras fails to be palmitoylated on the ER, it may be able to use the nonclassical pathway to move to the plasma membrane. Experiments with mono-palmitoylated H-Ras mutants are under way to test this prediction. 4 An additional prediction, suggested by the large amounts of H-Ras that associated stably with the plasma membrane when vesicular transport was inhibited, is that there are other palmitoyltransferases located at the plasma membrane. There is already some evidence that supports the possibility of H-Ras palmitoylation on the plasma membrane.  COS-7 cells were transfected with H-Ras Q61L DNA. Cycloheximide was added to the media immediately after the transfection and was removed 6 h after the transfection. Nocodazole (Noc) was added to the media upon the removal of cycloheximide, and after 2 h cycloheximide was added again to the media, whereas nocodazole was either washed out (ϪNoc) or still kept in the media (ϩNoc). After another 2 h, cells were fixed and immunostained. Arrows show the plasma membrane; arrowheads show the ERGIC membrane. Scale bar shows 10 m. AUGUST 31, 2007 • VOLUME 282 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 25767
novel H-Ras that is not farnesylated but that retains the native cysteines as sites for acylation completely avoids the Golgi, yet it is targeted to the plasma membrane and becomes palmitoylated (18,42).
In both yeast and as shown here in mammalian cells, H-Ras transport can function in the apparent absence of traditional vesicles. Neither interruption of outward vesicular movement with nocodazole nor ablation of COP I or COP II vesicles provided any evidence for nontraditional vesicles that might carry H-Ras to the plasma membrane. Thus, the data currently suggest that this nonclassical pathway is nonvesicular. The potential lack of a vesicular membrane for interaction of the lipidated C terminus of H-Ras poses a problem, as the hydrophobic farnesyl (and possibly, palmitoyl) group at the C terminus would then be exposed to the aqueous phase. This situation would be thermodynamically unfavorable unless the lipid groups were able to fold back against the protein. There is currently no precedent for such a model. An alternative way to prevent exposure of the lipids would be for an escort or chaperone protein to interact with the C terminus and mask the hydrophobic tail of H-Ras. There is also recent evidence that H-Ras can be found on intracellular particles termed "nanoparticles" or "rasosomes," which could potentially serve a role in H-Ras transport (43,44). Whatever the mechanism, this nonconventional transport of H-Ras is efficient and delivers H-Ras specifically to the plasma membrane.
Several cancer therapies currently in clinical trials are designed to inhibit membrane binding of Ras proteins (36). Our findings that H-Ras may use two pathways to reach the plasma membrane suggests that there is more to learn before this therapeutic strategy will be successful. Of equal importance, the evidence that a significant amount of H-Ras may use nonvesicular means for movement suggests that proteins that escort or regulate H-Ras trafficking from the ER should be sought. The assays developed in this work provide strategies that will be useful for searching for novel H-Ras partners.