Role of Xklp3, a Subunit of the Xenopus Kinesin II Heterotrimeric Complex, in Membrane Transport between the Endoplasmic Reticulum and the Golgi Apparatus

Partial Xklp3 cDNA clones were isolated from a Xenopus oocyte cDNA library (Vernos et al., 1993). DNA was sequenced on both strands using the T7, T3, and internal primers. The full-length cDNA for Xklp3 was reconstructed from three overlapping clones as follows: a Bluescript plasmid (Stratagene, La Jolla, CA) carrying the 5′ end sequence of Xklp3 cDNA was cut at a BclI internal site and EcoRI and ligated with a 730-bp internal fragment obtained by restriction digest of the second clone with BclI-EcoRI and a 3′ end fragment obtained from the third clone by digestion with EcoR1. The final construct was sequenced. Sequence analysis and comparisons were done with the DNASTAR program (London, UK) and coiled-coil predictions according to Lupas et al. (1991).

Fragments of Xklp3 corresponding to either the motor domain (Xklp3-M, nucleotides 217–1,248, amino acids 1–344) or part of the stalk and the full tail (Xklp3-T, nucleotides 1,684–2,446, amino acids 489–744) were generated by PCR using Pwo polymerase (Boehringer Mannheim, Indianapolis, IN) and inserted into the pGEX (Stratagene) and pMAL-c2 (New England Biolabs, Beverly, MA) vectors for expression in bacteria as recombinant proteins with the glutathione-S-transferase (GST) and the maltose-binding protein (MBP), respectively. The purified GST fusion proteins (Smith and Jonshon, 1988) were used to immunize rabbits following standard procedures (Harlow and Lane, 1988). Antibodies were affinity purified against the corresponding MBP fusion protein (Boleti et al., 1996) bound to CNBr-activated Sepharose (Pharmacia Biotech, Piscataway, NJ). The MBP-tail fusion protein (amino acids 489–744) was used to immunize mice and a monoclonal antibody (mAb X3T-3A6) was prepared following standard methods (Harlow and Lane, 1988).

A PCR fragment corresponding to the full stalk and tail domains of Xklp3 (amino acid 345–744) was generated by PCR using the Pwo polymerase, and cloned into the XhoI-BamH1 sites of pEGFP-C3 (Clontech, Palo Alto, CA) to generate pEGFP-Xklp3-ST. The construct HA-Xklp3-ST was generated by cutting out the green flourescent protein (GFP) epitope from pEGFP-Xklp3-ST with Nhe1-Sca1 restriction enzymes and replacing it with a 100-bp HA epitope generated by PCR from the vector pCMU-IV (Nilsson et al., 1989). The full-length GFP-Xklp1 was cloned in pEGFP-C3 and a fragment of the stalk and tail (nucleotides 2,370–3,740) was amplified by PCR and cloned in the vector pCDNA3.1 (Invitrogen, Carlsbad, CA) to generate Xklp1-ST-myc. The stalk region of GalT, GalNacT2, or Mannosidase II were fused to the NH₂ terminus of GFP in the following way. DNA fragments encoding the stalk regions were generated by PCR: (a) a 361-bp fragment (amino acids 1–120) from human GalT cDNA (GenBank/EMBL/DDBJ accession number X55415), (b) a 338-bp fragment (amino acids 1–114) from human GalNAcT2 cDNA (GenBank/ EMBL/DDBJ accession number X85019), and (c) a 348-bp fragment (amino acids 1–116) from mouse α-mannosidase II cDNA (GenBank/ EMBL/DDBJ accession number X61172). The PCR fragments were digested and cloned in pEGFP–N1 (Clontech) to generate pGalT–GFP, pGalNacT2–GFP, and pMannII–GFP. Vesicular stomatitus virus glycoprotein (VSV-G)–GFP was constructed by fusing GFP to the COOH terminus of the full-length VSV-G ts045 protein. A DNA fragment encoding the VSV-G ts045 protein (GenBank/EMBL/DDBJ accession number M11048) was generated by PCR and cloned in pEGFP-N1 to generate pVSV-G–GFP. All constructs were sequenced and did not contain any mutation altering the amino acid sequence.

Two frog cell lines were used, XL177 (Miller and Daniel, 1977), and A6 (Rafferty and Sherwin, 1969). Cells were grown at 23°C in L-15 medium with glutamine and 15% FCS. For transfection, A6 confluent cells were split 1:5 onto coverslips placed in 24-well plates, at least 20 h before the experiment. Cells were transfected with the Superfect reagent (QIAGEN, Santa Clarita, CA) according to the manufacturer's instructions. In short, 5 μg of plasmid DNA (Maxiprep; Qiagen) were used and the Superfect-DNA complex was left on the cells for 3 h before washing twice with 70% PBS and then twice with complete medium. Cells were grown for an additional 12–20 h before observation.

Stable cell lines expressing Golgi markers were generated by transfecting A6 cells at 30% confluency on a 10-cm Petri dish with 10 μg of DNA. Transfected cells were selected by adding 1 mg/ml of G418 (Geneticin; GIBCO BRL Life Tech, Paisley, Scotland, UK) to the culture medium. The GFP-positive cells were selected after 2 wk by flow cytometry on a FACSort^® (Becton Dickinson, Palo Alto, CA).

Rhodamine- or FITC-conjugated lectin from Helix pomatia (Sigma Chemical Co., St. Louis, MO) were resuspended in PBS and kept at 4°C. Nocodazole, cycloheximide, and BFA (Sigma Chemical Co.) were resuspended in DMSO, water, or methanol respectively, and kept at −20°C. Nocodazole was used at a final concentration of 5 μM and BFA at 5 μg/ ml. The following antibodies were used: a mouse monoclonal anti-protein disulfide isomerase (PDI) directed against the KDEL motif of the PDI protein, from S. Fuller (European Molecular Biological Laboratory [EMBL], Heidelberg, Germany); a sheep polyclonal anti–KDEL-receptor from J. Füllekrug, (EMBL) and I. Majoul, (Max Planck Institute, Göttingen, Germany); a polyclonal anti-GFP from K. Sawin (Imperial Cancer Research Fund, London, UK); a rabbit polyclonal anti-HA, from T. Nilsson (EMBL); a monoclonal anti-KRP85 and a polyclonal anti KAP-115 from J. Scholey, (University of California, Davis, CA); a monoclonal anti– α-tubulin purchased from Sigma Chemical Co. FITC-, Texas red-, cy3-, AMCA-conjugated secondary antibodies specific for rabbit or mouse were purchased from Dianova (Hamburg, Germany). Gold-coupled protein A was purchased from J.W. Slot, (Utrecht University, Utrecht, The Netherlands).

For immunofluorescence, fully confluent cells were split 1:2 onto coverslips and allowed to grow 24 h before fixation. Cells were washed in 70% PBS, fixed either in methanol at −20°C for 10 min or with glutaraldehyde as described in Boleti et al., (1997). The antibodies were diluted in PBS with 2% bovine serum albumin (BSA) and 0.1% Triton X-100. Primary antibodies were applied for 20 min at room temperature. The coverslips were washed three times for 5 min and secondary antibodies were then applied for 15 min. The coverslips were washed three times for 5 min and mounted in Mowiol (Hoechst, Frankfurt, Germany).

Confocal images were acquired using a confocal laser microscope Leica TCS-NT (Deerfield, IL) equipped with an Ar/Kr laser triple line. Confocal series were collected and projections were made using the TCS-NT software. For triple labeling experiments, we used a LSM-510 (Carl Zeiss Inc., Thornwood, NY) confocal microscope equipped with a Ar/Kr and a UV laser. In that case, single planes were collected. Image analysis was performed using Adobe Photoshop 4.0 (Adobe Systems, Mountain View, CA).

Cells were grown on a 10-cm Petri dish and fixed by adding to the medium an equivalent volume of 4% paraformaldehyde in 0.2 M sodium phosphate buffer, pH 7.4, for 1 h. The medium was then replaced with a fresh solution of 2% paraformaldehyde in 0.1 M sodium phosphate buffer and incubated for one more hour. Cells were collected by careful scraping and processed for cryosectioning according to Slot et al., (1991). The cryosections were retrieved with a 1:1 solution of 2.3 M sucrose and 2% methylcellulose according to Liou et al. (1996). Double immunogold labelings were performed as described previously (Slot et al., 1991). The monoclonal antibody X3T-3A6 was used in combination with a rabbit anti– mouse linker antibody, and then 10 nm of protein A–gold. The polyclonal anti–Xklp3-tail antibody was combined with 10 nm of protein A–gold. The polyclonal anti–KDEL-receptor or the polyclonal anti-GFP were combined with 5 nm of protein A–gold. Sections were stained, embedded in 2% methylcellulose/ 0.3% uranyl acetate, and then viewed in a Phillips CM120 electron microscope (Eindhoven, The Netherlands) at 80 KV.

The labeling densities of Xklp3 over the KDEL-receptor structures, the Golgi stacks (GalT GFP-positive), and the ER (PDI-positive) and the cytoplasm, were determined by the point-hit method (Weibel, 1979). 20 images were analyzed per marker. The KDEL-receptor compartment was defined as the tubular-vesicular structures areas marked by the erd-2 antibody, the Golgi stacks as membrane structures containing three or more cisternae and positive for the GalT–GFP, and the ER structures as the membranes marked by the anti-PDI antibody. The cross-section surface area for each compartment was measured using a square-lattice grid with a spacing d = 12.5 mm to count the points P corresponding to the grid line intersections comprised within the boundaries of the respective compartment. The surface area is then given by ΣPxd²/mag² (in μm²). The labeling density of Xklp3 in each compartment was calculated by dividing the number of gold particles that fall into the boundaries of each type of structure by the respective surface area. To evaluate Xklp3 density in the cytoplasm, we positioned a box corresponding to 10 points (d = 12.5 mm) on randomly selected areas of cytoplasm deprived of any membrane structures. We are conscious that this method might overestimate Xklp3 density in the cytoplasm, and for this reason, we counted a greater number of micrographs (n = 30). Background labeling was calculated by counting the number of gold particles detected on the mitochondria. It was usually ∼1 gold particle per μm² both for the polyclonal anti–Xklp3-tail and the mab-X3-T-3A6.

Interphase or cytosolic factor-arrested extracts from Xenopus eggs were prepared as described by Murray (1991). Motor proteins were prepared from egg extracts as described by Vernos et al. (1995). Whole cell extracts were prepared according to Boleti et al. (1996). 10 μg of anti–Xklp3-tail antibody coupled to 25 μl of Affiprep Protein A beads (Bio-Rad) were used for immunoprecipitation from 100 μl of egg extract (Walczak et al., 1996). Immunoprecipitations from transfected cells were performed in the following way. One 10-cm Petri dish of cells transfected 20 h before the experiment was incubated on ice with lysis buffer (25 mM Tris-Cl, pH 7.6, 200 mM NaCl, 1% Triton X-100, protease inhibitors). Cells were collected by scraping and spun at 12,000 g for 10 min. The supernatant was incubated for 1 h at 4°C with anti–GFP-coupled protein A–Sepharose (Pharmacia). The immunoprecipitate was washed three times (25 mM Tris-Cl pH 7.6, 200 mM Nacl, 0.1% Triton X-100) and resuspended in sample buffer before loading on SDS-PAGE and transfer on nitrocellulose for Western blot analysis.

Biochemical fractionations were done in the following way. A 10,000 g interphase egg extract was diluted three times with acetate buffer (Reinsch and Karsenti, 1997) and spun at 100 Kg for 30 min to obtain a membrane and a cytosol fraction. The membranes were resuspended in 2 M of sucrose in acetate buffer and loaded at the bottom of a sucrose step gradient (1.3/0.86/0.5/0.25 M sucrose in acetate buffer). The gradient was centrifuged overnight in a SW60 rotor (Beckman Instrs., Palo Alto, CA) at 300 Kg. The four layers were collected, centrifuged through a sucrose cushion, and then the pellets resuspended in one-tenth of the initial volume.

Cytosol fractionation from cells was performed in the following way. Cells grown on six 500 cm² plates were washed, scraped in 70% PBS, and then incubated in cytochalasin D. After centrifugation they were resuspended in KHEM (35 mM KCl, 50 mM Hepes, 5 mM EGTA, 1 mM MgCl₂, 3 μg/ml cycloheximide and 1 μM DTT, and protease inhibitors). Cells were broken by 10–15 passages through a ball-bearing homogenizer (8.0004-μm clearance) and centrifuged at 2,000 g to remove nuclei, heavy mitochondria, and unbroken cells. The postnuclear supernatant (PNS) was loaded at the top of a sucrose step gradient (0.25/0.5/0.86/1.3 M in KHEM) in SW40 tubes, and then centrifuged for 4 h at 200 Kg. 1-ml fractions were collected. The proteins contained in 200 μl of each fraction were precipitated with methanol/chloroform and loaded on SDS-PAGE followed by immunoblotting as described previously (Vernos et al., 1996).

Using a PCR-based approach, four kinesin-like proteins (Xklp1–Xklp4) were previously identified in Xenopus eggs (Vernos et al., 1993). We screened a Xenopus oocyte cDNA library with a probe corresponding to the motor domain of Xklp3. We obtained a 3,762-bp sequence from three overlapping clones containing an initiation codon at position 215, a stop codon at position 2,447, and a polyadenylation signal at position 3,716. The 2232-bp open reading frame encoded a protein of 744 residues with a calculated M_r of 85 kD.

The deduced protein sequence showed that Xklp3 had a conventional kinesin-like protein (KLP) organization with the NH₂-terminal motor domain, an α-helical region predicted to be involved in coiled-coil interactions and a globular COOH-terminal tail domain (Fig. 1,A). Alignment of the Xklp3 motor domain sequence with other KLP sequences had previously shown that Xklp3 was closely related to Drosophila KLP4 and KLP5 (renamed KLP64D and KLP68D), both members of the KIF3/KRP85/95 or Kinesin II subfamily (Vernos et al., 1993; Pesavento et al., 1994; Scholey, 1996). Xklp3 full-length sequence showed that it also shared a high degree of similarity to these KLPs outside the motor domain. The highest scores were obtained with KIF3B (Yamazaki et al., 1995), with 90% amino acid similarity in the motor domain, 85% in the stalk, and 74% in the tail (Fig. 1 B). This high level of conservation suggests that Xklp3 is probably the Xenopus counterpart of mouse KIF3B.

To characterize Xklp3 we raised two polyclonal antibodies: one against the motor domain (anti–Xklp3-M) and another against the COOH-terminal tail domain (anti– Xklp3-tail). Both affinity-purified antibodies, as well as a monoclonal antibody, mAb X3T-3A6, recognized a single band with an apparent M_r of 95 kD on immunoblots of extracts from Xenopus eggs and from two different Xenopus cell lines, XL177 and A6 (Fig. 2,A). We then looked at Xklp3 ATP-dependent microtubule-binding activity. In egg extracts, Xklp3 copelleted with endogenous taxol-stabilized microtubules in the presence of 5′-adenylyllimidodiphosphate (AMP-PNP) and was released from the microtubule pellet by 10 mM ATP (Fig. 2 B), similar to the behavior of kinesin and other KLPs (Bloom and Endow, 1994).

Most members of the Kinesin II subfamily form heterotrimeric complexes (Scholey, 1996). To determine whether this was also the case for Xklp3, we used the anti–Xklp3-tail antibody to immunoprecipitate the complex from Xenopus egg extracts. Three proteins of 95, 80, and 100 kD coimmunoprecipitated in equimolar amounts as estimated by SDS-PAGE gel (Fig. 2,C, arrows). Immunoblots of the immunoprecipitated material showed that the 95-kD band was Xklp3. The 80-kD band was recognized by an antibody against sea urchin KRP85 (the second motor subunit, Cole et al. [1993]) and the 100-kD band by an antibody against kinesin associated protein KAP115 (the third nonmotor subunit, Wedaman et al. [1996]). We conclude that Xklp3 is part of an heterotrimeric complex in Xenopus that is homologous to sea urchin Kinesin II and mouse KIF3A/KIF3B/KAP3 complexes (Fig. 1 B). Since in sea urchin and mouse, Kinesin II (Cole et al., 1993; Yamazaki et al., 1995) is a plus end–directed motor, it is likely that this is also the case for Xenopus Kinesin II.

We first determined the expression pattern of Xklp3 by Western blotting of different Xenopus tissues. We found Xklp3 in oocytes, eggs, testis, brain, XL177, and A6 cells. It was also present, although in lower amounts, in kidney and stomach and at very low levels in heart, lung, and muscle (data not shown). These data are consistent with the ubiquitous tissue distribution found for Kinesin II in mouse (Yamazaki et al., 1995).

We then examined the subcellular localization of Xklp3 in XL177 and A6 Xenopus cells. Immunofluorescence analysis of interphase XL177 cells with the anti–Xklp3-tail antibody revealed a bright staining concentrated in a perinuclear or juxtanuclear position (Fig. 2 E). The structure stained by the antibody appeared as a network of interconnected tubules reminiscent of the Golgi apparatus. In addition to this bright network, we could detect in some cells a light labeling of vesicle-like structures scattered throughout the cytoplasm. In mitotic cells, the labeled network disappeared and was replaced by a punctuate staining in the cytoplasm enriched around the spindle poles (data not shown). This staining pattern was similar to that observed in Xenopus dividing embryos (data not shown). We did not see any enrichment of labeled structures at the spindle equator in anaphase embryonic cells, as it has been reported for sea urchin Kinesin II (Henson et al., 1995).

To determine if the interphase network stained by the anti–Xklp3-tail antibody corresponded to the Golgi apparatus, we performed double immunofluorescence labeling on XL177 cells with the anti–Xklp3-tail antibody and different markers of Golgi membranes. Xklp3 staining colocalized extensively with the Helix pomatia (HP) lectin that binds N-acetyl-galactosamine residues generated at the first step of O-linked glycosylation (Pavelka and Ellinger, 1985; Roth, 1984), in the Golgi apparatus (Fig. 3,A). It colocalized perfectly with the KDEL-receptor (revealed by an anti– KDEL-receptor antibody, Fig. 3 A), an itinerant Golgi protein which recycles KDEL-containing ligands from the Golgi back to the ER (Lewis and Pelham, 1992). The KDEL-receptor is usually localized in the cis-Golgi and the intermediate compartment (Tang et al., 1993). We also observed a perfect colocalization of Xklp3 with β-COP a component of the COP I complex implicated in transport steps between the ER and the Golgi (data not shown).

To further define the localization of Xklp3 and because of the lack of cross-reacting antibodies, we prepared three stable lines of Xenopus A6 cells expressing GFP-tagged versions of Golgi resident enzymes: mannosidase II–GFP (MannII–GFP), galactosyl-N-acetyl transferase (GalNacT2– GFP), and galactosyltransferase (GalT–GFP). In mammalian cells it has been shown that addition of the GFP tag does not interfere with the targeting of these proteins to the Golgi apparatus (for review see Lippincott-Schwartz et al., 1998). We checked by immunofluorescence with FITC conjugated HP and the anti–KDEL-receptor antibody that all the GFP-tagged proteins localized to the Golgi network in Xenopus cells as well (data not shown). In addition, in vivo time-lapse confocal microscopy (data not shown) and EM (see below) also confirmed that the behavior of these proteins in Xenopus cells was similar to what has been described for mammalian cells (Sciaky et al., 1997; Shima et al., 1997). By immunofluorescence, Xklp3 colocalized with all GFP-tagged markers in the juxtanuclear network (GalNacT2–GFP, GalT–GFP [Fig. 3 B] and MannII–GFP [data not shown]); in addition, Xklp3 was present on dimly labeled vesicular structures throughout the cytoplasm.

To determine whether Xklp3 was also present in the ER, we stained cells with the anti–Xklp3-tail antibody and an anti-PDI, generated against the KDEL motif of lumenal ER resident proteins (Tooze et al., 1989). Anti–Xklp3-tail and anti-PDI antibodies did not stain the same network of membranes (Fig. 4). To examine whether Xklp3 was present on ER-to-Golgi or TGN to plasma membrane vesicles, we transiently transfected A6 cells with a VSV-G–GFP. VSV-G–GFP is exported from the ER to the Golgi apparatus where it is processed before being included in post-TGN membrane-bound structures and transported towards the plasma membrane (Arnheiter et al., 1984; Kreis and Lodish, 1986). Xklp3 and VSV-G–GFP colocalized extensively in the Golgi region, but no colocalization was seen in the peripheral (i.e., ER or post-TGN) vesicle-like structures or at plasma membrane extensions.

Altogether these results indicated that in A6 and XL177 Xenopus cell lines, Xklp3 was associated with Golgi membranes, but did not reveal to which specific Golgi subcompartment.

In a first attempt to determine whether Xklp3 was associated with Golgi stacks, we examined its relocalization after BFA treatment. BFA is used to discriminate secretory pathway subcompartments. This drug prevents the binding of peripheral coatomer proteins type 1 (COPI) proteins to Golgi membranes leading to membrane tubulation and redistribution of Golgi resident proteins into the ER (Doms et al., 1989; Scheel et al., 1997). However, some recycling proteins such as the KDEL-receptor, do not redistribute into the ER but are dispersed into vesicular structures throughout the cytoplasm (Tang et al., 1995; Fullekrüg et al., 1997). We verified that BFA treatment of Xenopus A6 and XL177 cells did result in a cytoplasmic redistribution of β-COP, a component of the COPI complex (data not shown). After 5 min of BFA treatment, Xklp3 was still present in some perinuclear structures and appeared in punctuate structures dispersed throughout the cytoplasm whereas the GalNacT2–GFP was associated with long tubular processes (Fig. 5,A). After 1 h of treatment, Xklp3 was associated with dispersed peripheral vesicular structures while GalNacT2–GFP had completely redistributed into the ER (Fig. 5,A). The lack of redistribution to the ER at later time points suggested that Xklp3 was not truly associated with the Golgi stacks. We then compared Xklp3 distribution to that of the KDEL receptor in BFA-treated cells (Fig. 5,B). During the first 30 min of BFA treatment, Xklp3 and the KDEL receptor colocalized on some vesicular structures (Fig. 5,B, 5 min, enlarged region). After 1 h of treatment, some of the vesicles stained by the anti-Xklp3 antibody were still positive for the KDEL receptor (Fig. 5 B, 1 hour, enlarged region) although a fraction of them were not. These results suggested that Xklp3 was associated with membrane structures distinct from the Golgi stacks and somehow related to the recycling compartment.

To identify the nature of the Golgi elements to which Xklp3 was associated, we examined thin frozen sections of cells using immunogold EM. Previous EM studies have shown that the Golgi apparatus is composed of stacks of flattened cisternae enriched in glycosylating enzymes, and tubulovesicular structures associated with the rims of the stacks probably involved in specific transport steps (Rambourg and Clermont, 1996). We first looked at the relative localization of Xklp3 and the GalT–GFP chimera using the monoclonal anti–Xklp3-tail (mAb X3T-3A6) and a polyclonal anti-GFP antibody followed by detection with protein A coupled to 10-nm (Xklp3) and 5-nm (GFP) gold, respectively. GalT–GFP was only found in the stacked cisternae of the Golgi complex as expected (Fig. 6,B, 5-nm gold), whereas Xklp3 was associated with tubular-vesicular structures sometimes in the proximity of the stacks (Fig. 6, A and B, arrows and arrowheads). A similar experiment using a polyclonal anti–KDEL-receptor and the same anti Xklp3 antibodies revealed that Xklp3 (10-nm gold) was found on tubulovesicular structures positive for the KDEL-receptor (Fig 6,A, arrows) (Table I, quantification). The rest of Xklp3 was localized partly in the cytoplasm or associated with unlabeled membranous elements some of which were morphologically similar to the ER. A double-labeling experiment with an anti-PDI antibody showed Xklp3 was localized as well on structures bearing PDI (Fig. 6,C, arrows) (Table I, quantification). EM studies have shown that the KDEL-receptor is distributed among the cis-Golgi, tubular-vesicular elements of the intermediate compartment and ER membranes (Griffiths et al., 1994). Thus the distribution of Xklp3 is similar to that of an intermediate compartment protein.

Membranes from the Golgi stacks, the ER, and recycling compartments behave differently upon sedimentation or floatation on sucrose gradients. We first separated the membranes from cytosolic proteins in egg extracts using a one step centrifugation, and further fractionated them by floatation. Approximately 20–25% of Xklp3 cosedimented with the membrane fraction (Fig. 7,A). We found Xklp3 associated with membranes that floated at the 1.3 M–0.86 M sucrose interface (like ER membranes) and at the 0.86 M–0.5 M sucrose interface (like Golgi membranes, Fig. 7,B). To further characterize the membrane fractions with which Xklp3 cosedimented, we prepared PNS from A6 cells stably transfected with MannII–GFP. The homogenates were then loaded on a sucrose step gradient (Fig. 7,C). Xklp3 was found in large amounts at the 0.25 M–0.5 M sucrose interface, probably corresponding to the soluble form of the trimeric complex. In addition, Xklp3 was found in two heavier fractions, at the 0.5 M–0.86 M sucrose interface which corresponded to a peak of Golgi membranes as revealed by the presence of MannII–GFP (Fig. 7 C, fraction 7, Xklp3, MannII–GFP) and at the 0.86 M–1.3 M sucrose interface which corresponded to a peak of ER/intermediate compartment as revealed by the presence of PDI. These results indicated that part of Xklp3 was associated with heavy membranes from both Golgi and ER/intermediate compartment.

The previous results pointed to a role for Xklp3 in the movement of tubular-vesicular elements between the ER and Golgi apparatus. To investigate the function of Xklp3, we transfected Xenopus cells with a plasmid expressing a truncated form of Xklp3 with either GFP (pEGFP– Xklp3-ST) or a hemagglutinin (HA) tag (pHA–Xklp3-ST) replacing the whole motor domain. Immunoprecipitation experiments with an anti-GFP antibody showed that the mutant protein could inhibit Kinesin II function by competing with the endogenous Xklp3 for the binding to the 85-kD subunit of the Kinesin II heterotrimeric complex (Fig. 1,B and Fig. 8,A). By immunofluorescence, both fusion proteins were mostly found distributed throughout the cytoplasm and not associated with the Golgi apparatus (Fig. 8, B and C). This could be explained in two ways: either the mutant protein could not be targeted to the vesicles to which the wild-type protein binds normally or the mutant protein prevented the localization of the Kinesin II complex to its target vesicles. In either case, this mutant clearly acted as a poison for the endogenous Kinesin II complex. Indeed, a similar construct has been used successfully to study Kinesin II dependent movement of the melanosomes in Xenopus melanophores (Tuma et al., 1998).

We first tried to detect an effect on one of the Golgi functions, glycosylation. The lectin from HP binds specifically to N-acetyl-galactosamine residues added at the beginning of the O-glycosylation. In A6 cells, it labels the Golgi apparatus, thus indicating that this step is occurring in the Golgi or in its direct vicinity. We stained the transfected cells with fluorescently labeled HP lectin and quantified the effect by scoring cells exhibiting a normal staining of the Golgi versus cells with no Golgi structure stained. Strikingly, the HP lectin no longer labeled the Golgi apparatus in 60–70% of Xklp3-ST–transfected cells (Fig. 8,C, asterisks; Fig. 8,D, quantification). This effect was similar in cells transfected with either tagged version of the mutant (Fig. 8,C). Transfected cells had a normal distribution of microtubules (MT) (Fig. 8,B), ER, mitochondria, and γ-tubulin (data not shown). As a control, we transfected cells with a plasmid expressing GFP-tagged Xklp1, a nuclear Xenopus kinesin-like protein involved in mitosis. Only 12% of the transfected cells had an abnormal HP lectin staining (Fig. 8,C, arrow; Fig. 8 D, quantification).

The effect of the Xklp3 mutant on lectin staining pattern suggested a defect in Golgi glycosylation function. The Xklp3 mutant could either affect directly the general structure of the Golgi or more specifically, the localization of the O-glycosylating enzymes. Alternatively, the mutant protein could affect the global transport of proteins from the ER to the Golgi.

To examine the morphology of the Golgi apparatus in the absence of Kinesin II function, we transfected cells stably expressing Golgi–GFP markers with the Xklp3 mutant. In these cells, the Golgi apparatus remained intact although the GFP staining was often weaker than in nontransfected cells (Fig. 8,E, asterisks). In the transfected cells, the HP lectin did not recognize any structure (Fig. 8,E, top) or very weakly the Golgi apparatus (Fig. 8,E, bottom). This suggested that the normal trafficking of proteins through the Golgi complex was altered. To test the hypothesis that the HP lectin detected N-acetylgalactosamine residues of cargo proteins en route along the secretory pathway during their transit through the Golgi apparatus, we blocked protein synthesis in cells expressing constitutively GalNacT2–GFP with cycloheximide to prevent the delivery of newly synthesized proteins to the Golgi apparatus. Under these conditions, the HP lectin did not label the Golgi apparatus although the Golgi was structurally intact as indicated by the GalNacT2–GFP fluorescence (Fig. 8 F). This result was very similar to the one we obtained in cells transfected with the mutant Xklp3 and suggested that the mutant form of Xklp3 was interfering with the delivery of newly synthesized proteins to the Golgi apparatus.

To examine whether the mutant form of Xklp3 blocked transport between the ER and the Golgi, we cotransfected cells with HA–Xklp3-ST and GalNacT2–GFP or GalT– GFP. The GFP markers localized exclusively to the Golgi in more than 80% of control cells. Cells expressing both HA–Xklp3-ST and one GFP Golgi marker showed a strong defect in the localization of the GFP marker to the Golgi (Fig. 9, A–C). Interestingly, the mislocalization of the Golgi GFP marker following co-transfection with the Xklp3-ST was associated with an aberrant HP lectin staining (Fig. 9,C). The phenotype observed could not be due to a general effect on the MTs because the cells still had a normal MT network focused at the MTOC (data not shown). As a control, we repeated the co-transfection experiment using a plasmid expressing a fragment of Xklp1 lacking the motor domain (Xklp1-ST–myc) similar to HA–Xklp3-ST. The Xklp1-ST–myc lacks also the NLS, and as a consequence remains in the cytoplasm, as HA– Xklp3-ST does. In cells expressing both Xklp1-ST–myc and GalNacT2–GFP, the GFP marker localized properly to the Golgi apparatus (Fig. 9 A).

Therefore, the results obtained with the dominant-negative form of Xklp3 indicate that Kinesin II is required for the proper transport and localization of newly synthesized proteins, to the Golgi apparatus.

In this study, we report the cloning and characterization of a subunit of the Xenopus heterotrimeric Kinesin II. Localization and functional data indicate that Xenopus Kinesin II plays an important role between the ER and the Golgi.

By immunofluorescence, we show that Xklp3 is associated with the Golgi apparatus. We see a extensive colocalization with GFP chimeras targeted to the stack or to the trans-face of the Golgi, as well as with the KDEL-receptor, a recycling molecule enriched on the cis-face of the Golgi. On the other hand, we find that Xklp3 does not colocalize with markers for the ER nor with VSV-G–labeled peripheral structures en route along the secretory pathway. Several experiments suggest that Xklp3 is actually present on membranes between the ER and the Golgi, on the so-called intermediate compartment (IC). First, upon treatment with BFA, Xklp3 remains associated with vesicles that disperse throughout the cytoplasm in a way similar to some IC markers (Tang et al., 1993; Griffith et al., 1994). Second, biochemical fractionation of the cytosol shows that Xklp3-associated membranes are found in fractions containing Golgi and ER markers. This behavior has been reported for several intermediate compartment proteins (Tang et al., 1993; Griffith et al., 1994). Third, by electron microscopy, Xklp3 is present mainly on tubular-vesicular structures that are sometimes closely associated with the Golgi stacks and are often stained by an anti– KDEL-receptor antibody. From all these data we conclude that, in Xenopus fibroblasts, Kinesin II is associated with a specific subset of membranes localized between the ER and the Golgi.

Using a mutant form of Xklp3 that functions as a dominant negative, we show that the Xenopus Kinesin II is involved in an essential aspect of Golgi function. When the mutant is overexpressed, the HP lectin no longer detects N-acetyl-galactosaminyl residues on proteins in the Golgi apparatus. This indicates that Xklp3 is required for the normal localization of the O-glycosylating enzymes and/or for the normal delivery of their substrates. In fact, Xklp3 is required for the localization of newly synthesized proteins to the Golgi apparatus. When the Xklp3 mutant is coexpressed with GFP–Golgi markers, these are mislocalized to the ER or to aggregates in the cytoplasm and do not reach the Golgi apparatus. Strikingly, in this case the staining of glycosilated residues by the HP lectin is also aberrant. This indicates that there is a correlation between the delivery of newly synthesized proteins to the Golgi and the defect in lectin staining. This interpretation is further supported by the finding that the inhibition of protein synthesis results in the absence of Golgi staining by the lectin. These results lead us to propose that Xklp3 function is required for the transport of newly synthesized proteins from the ER to the Golgi apparatus.

Given the association of Xklp3 with membranes localized between the ER and the Golgi apparatus, the effect of the mutant on the transport of proteins from the ER to the Golgi apparatus and the directionality of movement of Xklp3 along microtubules (plus end–directed), we propose two mechanisms of action for Xklp3 along the secretory pathway. We know that traffic between the ER and the Golgi apparatus involves an anterograde and a retrograde transport of vesicles which requires dynein and probably plus end–directed motors like kinesin. Dynein carries transport intermediates from the ER exit sites towards the MTOC where they integrate into the Golgi apparatus (Presley et al., 1997). Although this has not been demonstrated, it is possible that a plus end–directed motor counteracts the activity of dynein during anterograde transport contributing to the steady-state organization of the Golgi network. Indeed, using an antisense approach, Feiguin et al. (1994) have demonstrated that conventional kinesin may be needed for organizing the Golgi network in astrocytes. Because expressing the Xklp3 mutant interferes with the normal delivery of newly synthesized proteins to the Golgi, it is possible that Xklp3 participates in such a process.

On the other hand, two observations suggest that kinesin and/or KLP(s) are involved in the retrograde transport of vesicles from the Golgi apparatus to the ER. The inhibition of kinesin blocks the formation of tubular-vesicular structures between the Golgi and ER in the presence of BFA (Lippincott-Schwartz et al., 1995) and live studies have shown that the KDEL-receptor (important for the recycling of ER resident proteins) is associated with tubular-vesicular membrane structures forming at a velocity compatible with the involvement of a kinesin related protein (Sciaky et al., 1997). Therefore the steady-state size, location, and function of the Golgi apparatus may be strongly dependent on the relative rates of anterograde and retrograde transport. As a consequence, we can imagine that blocking a retrograde transport motor, here the Kinesin II, could also affect the anterograde pathway. An apparently surprising result is that even though inactivation of Xklp3 leads to a strong defect in ER to Golgi transport, it does not affect dramatically overall Golgi morphology, at least during the time course of our experiments. There are several possible explanations to this result. However, we favor the idea that blocking the retrograde transport results in freezing the flux of components throughout the secretory pathway, temporarily maintaining an apparently intact Golgi structure. It is also possible that by inhibiting Kinesin II, we block one specific part of the transport and that other motors keep other aspects of the secretory pathway active.

At this point, we emphasize that these two possibilities are only possible interpretations of our results, based both on our localization and functional data. More work will be required to establish exactly how the Xenopus Kinesin II is involved in the transport of components between the ER and the Golgi apparatus.

In various organisms, Kinesin II family members have been reported to associate with different types of membrane structures. In mouse neuronal cells, KIF3A/KIF3B/ KAP3 is found on vesicles, distinct from synaptic vesicles, in the cell body and along the axon (Yamazaki et al., 1995). It has been recently reported that mouse KIF3C (able to heterodimerize with KIF3A but not with KIF3B [the Xklp3 mouse homologue]) colocalizes with Giantin, a Golgi membrane protein, in spinal cord neuronal cells (Yang and Goldstein, 1998). In sea urchin embryos, Kinesin II is associated with vesicles enriched at the poles, and at the spindle interzone (Henson et al., 1995). Antibody microinjection experiments have shown that, in sea urchin, KRP95/KRP85/KAP115 is necessary during embryogenesis for correct delivery of ciliary components (Morris and Scholey, 1997). In Chlamydomonas, the fla10 gene product, the KRP85 homologue is associated to protein complexes beneath the flagellar membrane and mutant phenotypes suggest that FLA10 is involved in the intraflagellar transport (IFT) of protein complexes required for flagellar formation (Cole et al., 1998; Kozminski et al., 1995). In Xenopus, Kinesin II is present as well on purified melanosomes prepared from skin melanophores and a Xklp3 mutant protein equivalent to the Xklp3-ST used in this study inhibits the outward movement of melanosomes (Rogers et al., 1997; Tuma et al., 1998).

All these results indicate that Kinesin II associates with different organelles depending on the cell type and may play a role in a wide variety of processes. It is possible that this functional diversity is related to the heterotrimeric nature of the molecule and to the existence of isoforms or to the association of a core molecule with various proteins. Future work attempting to understand how it is targeted to a specific organelle and function should be of great interest.

It is worth noting that, in many cases, Kinesin II seems to antagonize the activity of dynein. In Chlamydomonas, cytoplasmic dynein is required to counteract FLA10 driven force and move the IFT particles back to the basal body of the flagella (Pazour et al., 1998). Xenopus melanosomes carry both the Kinesin II complex and cytoplasmic dynein. In this report we present evidence that Xenopus Kinesin II is involved in some step of the secretory pathway which also requires the activity of dynein.

This report is one of the first indications, together with the recent finding about the Rab6 interacting kinesin, and the mouse KIF1C, of the targeting of a KLP to a specific subcompartment of the secretory pathway and of a function for this KLP in this subcompartment. It is possible that different types of motors will be found targeted to specific domains of the secretory pathway, participating to the sorting process. The identification of such molecules may help to understand better the complexity of membrane traffic between the ER and the Golgi apparatus. The fact that the rabkinesin is targeted to the Golgi apparatus through a small G protein (Rab 6) raises the possibility that there is a whole pattern of localization of motors to specific membranes by adaptors like the G proteins. In this sense it is interesting to note that the human homologue of KAP3/KAP115, the third non motor subunit of Kinesin II, was cloned looking for proteins interacting with a small GTP-binding protein dissociation factor (Smg-GDS) (Shimizu et al., 1996). Smg GDS itself regulates the interaction of various small G proteins with membranes and might link the heterotrimeric kinesin to specific membrane cargoes via G protein interactions. Therefore, it will be of interest to examine whether different KLPs are involved in the transport of specific membrane compartments in the secretory pathway.

We would like to thank J. Füllekrug, S. Fuller, I. Majoul, T. Nilsson, K. Sawin, and J.M. Scholey, for the gift of antibodies. At EMBL, we would like to thank F. Senger and H. Wilhelm for their invaluable help in the production of antibodies, A. Atzberger for help with the cell sorter, and S. Röttger for help with the EM quantification. We also thank S. Reinsch, J. Mata, T. Nilsson, and T. Wittmann for discussions and advice, and J. Mata, T. Wittmann, and M. Zerial for critical reading of the manuscript.

AMP-PNP

5′-adenylyllimido-diphosphate

BFA

brefeldin A

GFP

green fluorescent protein

GST

glutathione-S-transferase

HA

hemagglutinin

HP

Helix pomatia

IC

intermediate compartment

IFT

intraflagellar transport

KLP

kinesin-like protein

MBP

maltose-binding protein

MT

microtubules

PDI

protein disulfide isomerase

PNS

postnuclear supernatant