A Novel Dynamin-like Protein Associates with Cytoplasmic Vesicles and Tubules of the Endoplasmic Reticulum in Mammalian Cells

Cell Culture and Tissues

Mouse hepatocytes (normal mouse liver cell line BNL CL.2; American Type Culture Collection [ATCC], Rockville, MD) and primary human foreskin fibroblasts were grown in MEM α medium with l-glutamine, ribonucleosides, and deoxyribonucleosides (GIBCO BRL, Gaithersburg, MD) containing 10% fetal bovine serum. The cultured normal rat liver cell line clone 9 (ATCC CRL-1439) was maintained in Ham's F-12K medium containing 10% fetal bovine serum. The cultured human cholangiocyte cell line (H-69; Grubman et al., 1994) was provided by Dr. Nicholas F. LaRusso (Mayo Clinic, Rochester, MN). Tissues used for this study were harvested from adult male Sprague-Dawley rats (Harlan, Madison, WI).

Antibodies

Antidynamin antibodies MC12 and MC63 were prepared as described previously (Henley and McNiven, 1996). For anti-DLP1 antibodies, three DLP1-specific peptides were selected: aa 266–294 for DLP-N (NNKKSVTDSIRDEYAFLQKKYPSLANRNG), aa 523–545 for DLP-MID (NNIEEQRRNRLARELPSAVSRDK), and aa 716–738 for DLP-C (DDLLTESEDMAQRRKEAADMLKA). These peptides were synthesized and conjugated to keyhole limpet hemocyanin, injected into New Zealand white rabbits, and then the antisera was collected. The crude antisera were affinity purified using corresponding HPLC-purified peptides immobilized on agarose columns. Polyclonal antibody to Rab1 was raised against a specific synthetic peptide, ATAGGAEKSNVKIQSTPVKQSGGGCC, in our laboratory and affinity purified as above. Polyclonal antibodies to Rab5 and TGN38 were prepared in our laboratory using the same methods. Mouse monoclonal anti–α-adaptin antibody (clone 100/2) was purchased from Sigma Chemical Co. (St. Louis, MO), anti–α-tubulin antibody from Amersham Corp. (Arlington Heights, IL), and anticalnexin antibody from Affinity BioReagents, Inc. (Golden, CO). Mouse monoclonal anti-clathrin heavy chain antibody (X22) was prepared from culture supernatant of a hybridoma cell line (ATCC CRL-2228). Mouse monoclonal antidynamin antibody (hudy-1) was provided by Dr. S.L. Schmid (Scripps Research Institute, La Jolla, CA). The following rabbit polyclonal antibodies were provided as gifts: anti–γ-adaptin from Dr. L. Traub (Washington University, St. Louis, MO), anti–β-galactosidase from Dr. N. LaRusso (Mayo Clinic, Rochester, MN), anti-Sec23p from Dr. J.-P. Paccaud (University of Geneva, Geneva, Switzerland), anti-Rab8 from Dr. D.D. Sabatini (New York University, NY), anti-p47 (μ3) from Dr. M.S. Robinson (University of Cambridge, Cambridge, UK). For secondary antibodies: FITC-, TRITC- (Kirkegaard and Perry Laboratories, Gaithersburg, MD), or Texas red-conjugated (Molecular Probes Inc., Eugene, OR) goat anti–rabbit and goat anti–mouse IgG and Cy3-conjugated donkey anti–rabbit IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) were used for immunocytochemistry. HRP-conjugated goat anti–rabbit and goat anti–mouse IgG (Biosource International, Camarillo, CA) were used for Western blotting.

Isolation of DLP1 and Peptide Microsequencing

To isolate DLP1, a large scale immunoprecipitation was performed using the antidynamin antibody MC12 from rat liver homogenate according to previously described methods (Henley and McNiven, 1996). The immunoprecipitates were run on 5–15% gradient SDS-PAGE and proteins were visualized as described by Rosenfeld et al. (1992). A protein band corresponding to the 80-kD protein was cut out of the gel. Tryptic digestion, peptide elution, and microsequencing were performed by the Mayo Protein Core Facility.

RT-PCR

Total RNA was isolated from rat tissues by guanidium thiocyanate-phenol-chloroform treatment followed by isopropanol precipitation (Chomczynski and Sacchi, 1987). A cDNA was prepared with Moloney murine leukemia virus reverse transcriptase (New England Biolabs Inc., Beverly, MA) and random primers. For DLP1 cloning, PCR was performed with cDNA as a template for 35 cycles at 95°C for 1 min, 52°C for 2 min, and 72°C for 1 min. Degenerate primers were 5′-AGYTCRGTGCTSGARAGYTTRGTNGG-3′ (256-fold) and 5′-GAYTTYGAYGARATHMGNCARGARATH-3′ (4,608-fold) for forward primers and 5′-TTNACCACRTTKGGNGARAANAC-3′ (512-fold) for the reverse primer. For the PCR at the alternative splicing regions, reactions were carried out for 25 cycles at 95°C for 1 min, 56°C for 2 min, and 72°C for 1 min with primers 5′-CTCAGTGCTGGAAAGCCTAGT-3′ and 5′-CATGATGCCATAGTTGAAGTA-3′ for forward primers and 5′-TTTACCCCATTCTTCTGCTTC-3′ and 5′-AATTCCACCACCTGCAGATGC-3′ for reverse primers.

Cloning and Sequencing

Rat brain and liver λ-ZAPII cDNA libraries (Stratagene, La Jolla, CA) were screened using a ³²P-labeled DNA probe that was obtained by reverse transcriptase (RT)–PCR. Recombinant plaques were transferred onto nitrocellulose filters (Schleicher & Schuell, Inc., Keene, NH) and cross-linked at 120 mJ/cm² in a UV cross-linker (Fisher Scientific Co., Pittsburgh, PA). Filters were prehybridized for 4 h at 65°C in 6× SSC containing 0.25% nonfat dried milk, then hybridized overnight in the same solution containing random-primed ³²P-labeled DNA probes. Washing was performed with a final stringency of 0.2× SSC/0.02% SDS at 68°C. Positive plaques were isolated and further purified by secondary and tertiary screenings under the same conditions used in the primary screening. Positive clones were rescued and maintained as plasmids (rat DNA in pBluescript SK−) with helper phage (ExAssist; Stratagene), according to the manufacturer's manual. DNA sequencing was performed by an automated DNA sequencer in the Mayo Molecular Biology Core Facility. Sequence comparisons and other sequence analyses were performed using DNASTAR sequence analysis software (DNASTAR Inc., Madison, WI).

Indirect Immunofluorescence Microscopy

Immunofluorescence microscopy was performed as described in previous publications (Marks et al., 1995; Henley and McNiven, 1996). To double stain cells with fluorescently labeled dextran, cells were incubated with medium containing 100 mM lysine-fixable FITC-conjugated dextran (3,000 MW; Molecular Probes, Inc.) for 1.5 h at 37°C, acid washed for 1 min with cold HBSS (Sigma Chemical Co.) adjusted to pH 3.0 with acetic acid, and then rinsed with five changes of cold normal HBSS (pH 7.2). Cells were then fixed and processed for indirect immunofluorescence microscopy. To label endosomes containing transferrin, cells were incubated in serum-free medium (DME plus 0.02% BSA) for 30 min before incubation with 5 μg/ml FITC-conjugated human transferrin (Molecular Probes, Inc.) for 15 min at 37°C. Rinsing and processing for indirect immunofluorescence were the same as in dextran labeling. For immunofluorescence staining with cells expressing GFP-DLP1, cells were prepermeabilized with 0.01% digitonin for 30 s in microtubule stabilizing buffer (100 mM Pipes, pH 6.9, 5 mM MgSO₄, 1 mM EGTA, 20% glycerol) and processed for standard indirect immunofluorescence microscopy.

Immunoelectron Microscopy

All steps were carried out essentially as described previously (Dahan et al., 1994). In brief, overnight fasted male or female Sprague-Dawley rats were anesthetized with Somnotol. Livers were perfusion fixed via the portal vein with 4% paraformaldehyde/0.5% glutaraldehyde in 0.1 M PO₄ buffer (pH 7.5) for 10 min. Small 1-mm³ pieces of liver were dissected out and left in the fixative at 4°C for 1 h and then washed in 4–5 changes of ice-cold 4% sucrose/0.1 M PO₄ buffer. Tissues samples were then cryoprotected in 2.3 M sucrose/0.1 M PO₄, mounted on nickel stubs, and quick-frozen in liquid nitrogen. Cryosections were cut and labeled the same day as described previously (Dahan et al., 1994). Antibody dilutions were 1:2 for rabbit anti–rat DLP-N and 1:20 for the goat anti–rabbit secondary gold (10 nm) conjugate. After immunolabeling procedures, sections were stained with uranyl acetate oxalate, washed in two changes of water, and then transferred to drops of methyl cellulose containing 0.4% aqueous uranyl acetate on ice and then air dried. Sections were viewed in an electron microscope at 80 kV (Philips Electron Optics, Mahwah, NJ). Labeling density measurements were evaluated by dividing the number of gold particles over a particular structure by its profile area in μm² as determined on the measuring tablet of a Zeiss MOP-3 digitizer (Carl Zeiss Inc., Don Mills, Ontario).

Liver Subcellular Fractionation

Subcellular fractionation was performed as described by Fleischer and Kervina (1974) and illustrated as a simplified scheme shown in Diagram 1. In brief, rat livers were harvested and homogenized in 5 vol (vol/wt) of buffer H (0.25 M sucrose in 10 mM Hepes, pH 7.5). The homogenate was spun at 960 g for 10 min. The supernatant (S1) was saved to isolate Golgi fraction, microsomes, and cytosol. For the fractionation of nuclei, mitochondria, and plasma membranes, the pellet (P1) was resuspended to a final sucrose concentration of 1.6 M, overlaid with two-thirds vol of buffer H, and spun at 71,000 g for 70 min in a Beckman SW28 rotor (Beckman Instruments, Inc., Fullerton, CA). While the nuclear fraction was collected as the pellet, membranes enriched at the interface between 0.25 and 1.6 M sucrose were collected and washed twice. The sucrose concentration of resuspended membranes was adjusted to 1.45 M, layered under buffer H, and spun at 68,000 g for 60 min in a Beckman Ti70 rotor. Mitochondrial fractions were recovered as a pellet while the plasma membrane was enriched at the interface. For the fractionation of Golgi apparatus, microsomes, and cytosol, S1 was spun at 34,000 g for 10 min and the pellet was discarded. The supernatant (S2) was spun at 50,000 g for 30 min in the Beckman Ti70, and the resulting supernatant (S3) was spun again at 200,000 g for 60 min. The supernatant (S4) was collected as a cytosolic fraction and the pellet (P4) as the light microsomal fraction. P3 was resuspended gently using a homogenizer in 10 mM Hepes, pH 7.4, containing 52% sucrose, then the sucrose concentration was adjusted to 43.7%. Sucrose concentrations of 38.7, 36, 33, and 29% solutions were sequentially layered on top of the 43.7% sucrose, which contained membrane mixtures, and spun at 120,000 g for 53 min in a SW28 rotor. Golgi fractions were recovered from the 29 and 33% sucrose interface, and heavy microsomes were at the bottom of the gradient. To fractionate rough and smooth microsomes (RM and SM, respectively), equal portions of heavy and light microsomes were combined, adjusted to 0.25 M sucrose, and made to 0.015 M CsCl. The mixture was layered on top of 1.3 M sucrose containing 0.015 M CsCl and spun at 300,000 g for 110 min in a Beckman Ti70 rotor. SM were enriched at the interface and RM were collected as a pink sediment at the bottom.

Liver Microsome Fractionation

Rat liver microsomes were fractionated by methods described previously (Howell et al., 1978; Howell and Palade, 1982) except for the buffer composition. In this experiment, 50 mM imidazole, pH 7.4 and 250 mM sucrose were used for the initial homogenization and total microsome isolation. In brief, rat liver was homogenized and centrifuged at 10,000 g for 10 min to remove cell debris, nuclei, and mitochondria. Total microsomes were obtained by centrifuging the postmitochondrial supernatant at 100,000 g for 90 min. The total microsomes were resuspended, made to 1.22 M sucrose, and loaded under a sucrose step gradient of 1.15, 0.86, and 0.25 M sucrose. The gradient was centrifuged for 3 h at 82,500 g. Light Golgi fraction (GL) was obtained as a material floated to 0.25/0.86 M sucrose interface. Heavy Golgi fraction (GH) was a material at the 0.86/1.15 M sucrose interface, and ER fraction was recovered as a pellet of the centrifugation.

SDS-PAGE and Western and Northern Blot Analyses

SDS-PAGE was conducted using a buffer system based on the method of Laemmli (1970). Western blotting procedure is performed as described previously (Henley and McNiven, 1996) except using 10% nonfat dry milk as a blocking reagent. Northern blot analysis was essentially carried out using the methods described in a previous publication (Török et al., 1996).

Identification of a Novel Dynamin-like Protein

An antibody raised against a highly conserved GTP-binding region of dynamin consistently recognized a protein of 80 kD in addition to dynamin, as assessed by immunoblot analyses of a rat liver extract (Fig. 1). This 80-kD protein was isolated from liver homogenates by immunoprecipitation, and amino acid sequences of several tryptic peptides were determined. These sequences showed striking similarity to dynamin but indicated that the 80-kD protein was a novel protein distinct from dynamin (Fig. 1). Degenerate primers were designed based on the amino acid sequences of the peptides, and a DNA probe was generated by RT-PCR. Screening of a rat cDNA library identified a full-length cDNA coding for a novel dynamin-like protein, DLP1. DLP1 was encoded by an open reading frame of 755 aa with a calculated molecular mass of 83.9 kD (Fig. 2 a; Yoon, Y., and M.A. McNiven. 1996. Mol. Biol. Cell. 7:82a). While we were performing this study, we found that a partial sequence of the same protein from human had been deposited in the EST database (Adams et al., 1995). Amino acid sequence comparison of rat DLP1 with rat Dyn1 and two yeast dynamin-related proteins, Dnm1p and Vps1p, is shown in Fig. 2 a. As for all dynamin family members, there is a high degree of homology between DLP1 and the conventional dynamins within the NH₂-terminal domain. The tripartite GTP-binding motif of dynamin is highly conserved in DLP1, strongly indicating that DLP1 is a GTP-binding protein. Like Dnm1p and Vps1p, DLP1 does not have the proline-rich tail of dynamin. All four sequences are completely divergent at the region aligned with the pleckstrin homology (PH) domain of dynamin (aa 510–633 of Dyn1) in this sequence comparison. Homology is partially restored at the extreme COOH-terminal end of DLP1. Portions of this region were selected as having high probability of coiled coil formation by the sequence analysis program: aa 711–752 of DLP1, aa 654– 681 and 710–741 of Dyn1, and aa 666–699 of Vps1p.

Interestingly, DLP1 shares greater homology with the dynamin-related proteins from Saccharomyces cerevisiae, such as Dnm1p (42%) and Vps1p (40%), than with the conventional mammalian dynamins (36%; Fig. 2 b). Whether DLP1 can be considered a true mammalian homologue of Vps1p or Dnm1p is unclear at this time. According to the phylogenetic tree shown in Fig. 2 c, the dynamin family has been expanded and is subdivided into three groups. In addition to the original group of dynamins, including the Drosophila shibire gene product, a second group includes rat DLP1 described here and three yeast proteins (Dnm1p, Vps1p, and Schizosaccharomyces pombe dynamin-like protein). The third group consists of plant dynamin-like proteins (Dombrowski and Raikhel, 1995; Gu and Verma, 1996).

DLP1 Is Expressed in All Tissues as Multiple Alternatively Spliced Variants

Northern and Western blot analyses of various rat tissues showed that DLP1 is expressed ubiquitously. By Northern analysis (Fig. 3 a), all tissues examined expressed a 4.6-kb transcript. In testis, however, an additional 3-kb message was also detected. This smaller transcript would be large enough to encode a full-length DLP1 if a tissue-specific splicing event occurred outside the coding region. Alternatively, a truncated form of DLP1 may be encoded by this shorter transcript. Three different clones encoding shorter open reading frames were isolated from the library screening. All three were truncated by ∼200 aa at the COOH termini (results not shown). In preliminary studies using RT-PCR with primers specific for the truncated forms, only one of them was found to be expressed. However, this truncated form was expressed ubiquitously and at a much lower level than that of the full-length DLP1 (results not shown) suggesting that it is not encoded by the 3-kb message detected in testis.

To define the distribution of DLP1 protein in tissues and cells, peptides to sequences representing the middle, NH₂-terminal, and COOH-terminal regions of DLP1 were synthesized and specific antibodies were raised (see Materials and Methods). Immunoblot analyses of cell and tissue homogenates were conducted, and they confirmed the observations made from the Northern analyses that DLP1 is expressed in a wide variety of cell types and tissues (Fig. 3, b and c). Interestingly, brain DLP1 ran significantly higher on SDS-PAGE than DLP1 from other tissues (Fig. 3 b). It was also noted that the most prominent DLP1 band from both cultured mouse hepatocytes and human fibroblasts ran slightly lower than DLP1 from the rat nonneuronal tissues (Fig. 3 c). To test for the existence of alternatively spliced forms of DLP1 that could explain the multiple molecular mass seen between brain and other tissues, additional DNA sequencing of positive clones was performed and revealed two alternatively spliced regions (Fig. 4). In each region, at least three differentially spliced forms were found. In addition, RT-PCR was performed with RNA from four different tissues using primers specific for the flanking sequences of two alternative splicing regions. The predicted sizes of the different splicing variants were 201, 180, and 162 bp at the NH₂-terminal splicing region and 351, 318, and 240 bp at the COOH-terminal region (Fig. 4). We found that brain preferentially expressed the longer forms at both regions, whereas other tissues expressed shorter forms almost exclusively (Fig. 4). RT-PCR with brain mRNA mainly amplified a 201-bp fragment at the NH₂-terminal splicing region and a 351-bp fragment at the COOH-terminal splicing region with some amplification of the shorter forms. RT-PCR with RNA from liver, lung, and testis predominantly amplified 162- and 240-bp fragments at NH₂-terminal and COOH-terminal splicing regions, respectively. Additional faint bands were detected by the RT-PCR, suggesting that there may be more splicing variants in these regions. The calculated molecular mass of the brain DLP1 form is 83.9 kD and that of other tissues is 78.6 kD.

Subcellular Localization of DLP1

To identify the cytoplasmic compartment with which DLP1 associates, indirect immunofluorescence microscopy was performed on a cultured rat hepatocyte cell line (clone 9) and human fibroblasts using the affinity-purified, polyclonal peptide antibodies described above. As shown in Fig. 5, antibodies to DLP1 stain punctate vesicular structures that are concentrated at the perinuclear region and extend out into the peripheral cytoplasm in linear arrays as if associated with cytoskeletal elements. Double fluorescence labeling of cells for DLP1 and endocytosed fluorescent dextran or transferrin did not reveal a striking colocalization (Fig. 6 a), suggesting that DLP1 is not associated with endocytic or lysosomal compartments.

Immunoblot analyses of rat liver subcellular membrane fractions supported the immunofluorescence observations and suggested that DLP1 is associated with small nonendocytic organelles. Although a large portion of DLP1 was found to be cytosolic, a substantial amount was associated with a light microsomal fraction containing light membranous cytoplasmic vesicles (Fig. 6, b and c). To characterize this fraction, we immunoblotted subcellular fractions with antibodies to endocytic or lysosomal compartments. The endocytic markers dynamin and α-adaptin were enriched in the plasma membrane fraction as predicted, whereas Rab5 and the lysosomal marker β-galactosidase was found in multiple fractions (Fig. 6 b). Importantly, conventional dynamin was enriched in both plasma membrane and Golgi membrane fractions with little found in the DLP1-containing light microsomal fraction.

Because the distribution of DLP1 did not coincide significantly with endocytic compartments, we next probed liver subcellular fractions with markers for the secretory pathway. The markers used include antibodies to γ-adaptin, a subunit of the AP-1 adaptor complex that participates in TGN-derived, clathrin-coated vesicle formation (Ahle et al., 1988; Robinson, 1992); COPII (Sec23p), a marker for ER-to-Golgi transport vesicles (Barlowe et al., 1994); Rab8, a marker for a subpopulation of post-Golgi secretory vesicles (Huber et al., 1993); and μ3, a subunit of the recently identified adaptor complex AP-3 that may be involved in nonclathrin-coated vesicle formation from the TGN (Simpson et al., 1996; Dell'Angelica et al., 1997; Robinson, 1997). As shown in Fig. 6 c, multiple secretory membrane proteins were significantly enriched in the DLP1 light microsomal fraction. The coenrichment of these secretory vesicle proteins with DLP1 in the light microsomal fraction suggests that DLP1 may associate with a secretory compartment.

It is possible, although unlikely, based on the immunofluorescence images, that DLP1 forms large protein aggregates and pellets with membranes as opposed to being vesicle-associated in the differential centrifugation. To address this concern, we performed a second, well-characterized density gradient centrifugation method in which vesicles float upward in the sucrose gradient as opposed to pelleting (Ehrenreich et al., 1973; Howell et al., 1978; Howell and Palade, 1982). With this method, we found that, in addition to being cytosolic, DLP1 was enriched in light Golgi and ER fractions (Fig. 6 d). The light Golgi fraction, which contains multiple small vesicles, was obtained by floatation of total liver microsomes from the 1.22 M sucrose load (see Materials and Methods), indicating that DLP1 is associated with vesicle membranes. To further verify the cofractionation of DLP1 with the ER, an extended fractionation of microsomal membranes was conducted by the method used in Fig. 6, b and c. Immunoblotting of these fractions again revealed DLP1 enrichment in the rough microsomal fraction (RM) where ER was enriched (Fig. 6 e). In this fractionation, Rab1, a marker for ER-to-Golgi transport, was enriched in Golgi, smooth ER (SM), and rough ER (RM), and ribophorin II, a resident ER protein in SM and RM (results not shown) as previously shown (Plutner et al., 1991).

In the experiments described above, we have applied two different subcellular fractionation methods and found that DLP1 is enriched in ER fraction and a light vesicle membrane compartment as compared with dynamin that is enriched in plasma membrane and Golgi fractions. To confirm that dynamin and DLP1 are on different membrane compartments, we performed double label immunofluorescence microscopy of cultured hepatocytes with antibodies to DLP1, dynamin, and clathrin. As shown in Fig. 7 a, there is no significant colocalization between DLP1 and dynamin. Whereas DLP1 antibodies stained smaller vesicular structures concentrated in the perinuclear region, dynamin antibodies (hudy-1) recognized larger punctate structures, presumably clathrin-coated pits on the plasma membrane (Damke et al., 1994). Additional double label immunofluorescence microscopy using DLP1 and clathrin heavy chain antibodies showed that DLP1 is not localized to the clathrin-containing organelles that are present at the TGN and the plasma membrane (Fig. 7 b). These results, in combination with the immunoblot analyses of subcellular fractions, suggest that DLP1 is not associated with an endocytic compartment and has a localization distinct from that of conventional dynamin.

DLP1-positive Structures Associate with Microtubules and Tubules of the Endoplasmic Reticulum

The immunofluorescence micrographs of cellular DLP1 shown in Fig. 5 suggested that vesicles that associate with DLP1 aligned along linear structures reminiscent of microtubule arrays emanating from a perinuclear centrosome. To further define this linear organization, double immunofluorescence microscopy was performed using antibodies to DLP1 and tubulin. As shown in Fig. 8, a large portion of DLP1-positive vesicles coincided with microtubules. This association is most striking in the color double-stained images (Fig. 8, c and inset) that reveal discrete DLP1-positive vesicles bound along single microtubules.

Although we observed significant overlap between the DLP1 and microtubules, it was unclear as to why long prominent chains of DLP1-positive structures appeared to align along some microtubule strands but not others. This suggested that the DLP1 vesicles might selectively associate with a membrane organelle that extends along the microtubule proper. Based on the fact that DLP1 is enriched in ER containing subcellular fractions as opposed to endocytic compartments (Fig. 6), and that ER tubules are known to align along microtubule networks (Terasaki et al., 1986; Lee et al., 1989), we conducted double immunofluorescence staining for DLP1 and two ER markers, calnexin and Rab 1. Calnexin is a resident chaperone protein of the ER (Rajagopalan et al., 1994) while Rab1 has been implicated in regulation of vesicle transport between the ER and the Golgi apparatus (Plutner et al., 1991; Tisdale et al., 1992; Davidson and Balch, 1993; Nuoffer et al., 1994; Pind et al., 1994). As shown in Fig. 9, we observed a remarkable overlap between DLP1 and calnexin-positive ER tubules. Although DLP1 vesicles were observed throughout the cytoplasm, the coalignment between the ER and the long extended arrays of DLP1-positive structures was nearly absolute and very striking. In high magnification images (Fig. 9, inset), one can observe many DLP1 punctate spots situated along a single ER cisterna. Next, we compared the DLP1 distribution to that of Rab1. Because our antibodies to both Rab1 and DLP1 were made in rabbits making double staining difficult, we used clone 9 cells stably transfected to express DLP1 fused to the green fluorescent protein (GFP-DLP1; K.R. Pitts, Y. Yoon, and M.A. McNiven, manuscript in preparation). Because GFP-DLP1 in these cells was overexpressed and filled the cytoplasm, the cells were permeabilized before fixation to allow cytosol and many of the individual vesicles to leak out of the cell, resulting in a reduced cytoplasmic fluorescence. After permeabilization, fixation, and immunostaining for Rab1 of the transfected cells, we observed prominent linear distributions of GFP-DLP1 spots that colocalized with the Rab1-positive compartment (Fig. 10). Since the Rab1 has been shown to distribute at ER as well as Golgi regions (Plutner et al., 1991) and GFP-DLP1 did not localize with Golgi markers by immunofluorescence (results not shown), the Rab1-positive GFP-DLP1 tubules are ER cisternae, confirming DLP1 localization at ER tubules. Interestingly, the two proteins appear to coalign with each other but do not overlap (Fig. 10, c and d). Thus, one can observe a consistent alternating pattern of Rab1- and DLP1-positive structures.

As a final confirmation of DLP1 localization to ER tubules, immunogold labeling of hepatocytes was performed on ultrathin frozen sections of rat liver. Fig. 11 is a collection of representative images showing a prominent labeling of ER cisternae and tubular networks by DLP1 antibodies as compared with other organelles. There was little or no labeling of Golgi stacks proper but some gold particles were found on closely associated tubular networks. Table I provides a quantitation of gold particles on different cytoplasmic organelles, revealing that DLP1 is predominantly found along tubular reticular membranes of ER. Golgi stacks are depleted in DLP1 immunoreactivity as revealed by a labeling density similar to background. Thus, taken together, the biochemical and morphological observations reported here are consistent with the premise that DLP1 is associated with both ER tubules and cytoplasmic vesicles.

A Novel Dynamin-like Protein in Mammalian Cells

In this study, we have identified a novel dynamin-like protein, DLP1, that shares 34–37% and 40–42% overall homology to conventional dynamins and yeast dynamin- related proteins, respectively (Yoon, Y., and M.A. McNiven. 1996. Mol. Biol. Cell. 7:82a). According to a sequence comparison of DLP1 with dynamin, Vps1p, and Dnm1p, the NH₂-terminal enzymatic domains of these proteins are highly conserved except for a stretch of 25–45 aa between the first and second GTP-binding elements of DLP1, Dnm1p, and Vps1p. This amino acid cassette is not present in dynamin and contains the first of two alternatively spliced sites of DLP1 that will be discussed later. Sequence homology among these proteins is very low at the region aligned with the dynamin PH domain (aa 510–633 of Dyn1) in the comparison shown in Fig. 2 a. The PH domain has been suggested to mediate membrane binding of dynamin and other PH domain–containing proteins (Lemmon et al., 1996; Salim et al., 1996). It is unclear whether this region of DLP1 forms a PH domain because it cannot be predicted by primary amino acid sequence comparison (Downing et al., 1994; Ferguson et al., 1994; Lemmon et al., 1996). However, secondary structure predictions of both DLP1 and Dnm1p show similar features in this region to that of known PH domains that contain alternating structures of several β-sheets and variable loops (Ferguson et al., 1994). The divergent sequence in this region suggests that this domain may render different organelle specificity to the dynamins and dynamin-related proteins. Finally, the extreme COOH-terminal region of DLP1 shows a small but significant restoration of homology to both dynamin and the two yeast proteins. This tail region of DLP1 is predicted to form a coiled coil, suggesting that it may participate in the formation of polymeric structures with other proteins.

A Ubiquitous Dynamin-like Protein That Is Alternatively Spliced in a Tissue-specific Manner

Northern and Western blot analyses indicate that DLP1 is expressed ubiquitously, but a tissue-specific expression of different alternatively spliced variants exists. Two alternative splicing regions were found in DLP1. As mentioned, the first is in the amino acid cassette that is not present in the dynamins. The second is in the region aligned with the PH domain of dynamin. Both of these regions are highly divergent from the dynamins and the other dynamin- related proteins. Further sequence analysis indicates that these alternatively spliced regions have a high surface probability. Thus, these regions are likely to be exposed and may bind to membrane lipids or other proteins. Therefore, different splicing variants produced in these regions may have different binding properties. Such alterations in binding properties by alternative splicing have been reported for many proteins, such as agrin, in which a 4-aa insertion creates a specific heparin-binding site (Gesemann et al., 1996). Interestingly, DLP1 exhibits tissue-specific alternative splicing, with brain expressing a larger form that includes an additional 50 aa at two splicing regions (Figs. 3 b and 4). Whether these insertions in the brain form of DLP1 modify its distribution or function for rapid secretory or endocytic processes in neurons remains to be tested.

A Vesicle-associated Dynamin-like Protein with a Distribution Distinct from Conventional Dynamin and Endocytic Markers

We have analyzed the cytoplasmic distribution of DLP1 using multiple methods including immunoblotting of subcellular fractions (Fig. 6), immunofluorescence microscopy (Figs. 5, 7–9), the expression of GFP-tagged DLP1 in cultured cells (Fig. 10), and immunogold labeling (Fig. 11). By morphological criteria, antibodies to DLP1 clearly label a vesicular compartment. Further, GFP-DLP1 expressed in cultured cells shows a vesicular distribution that is near identical to that observed by antibody staining. In addition to this striking vesicular morphology, Western blot analysis of rat liver membrane fractions revealed that DLP1 is soluble and a significant portion is associated with a light membrane vesicle fraction (Fig. 6, b and c). To insure that this fractionation represents a true association with vesicles as opposed to protein aggregates, we used another sucrose gradient centrifugation method (Fig. 6 d). Regardless of the fractionation method implemented, we consistently observed DLP1 associated with lighter membranes or vesicles. In addition, DLP1-positive vesicles did not colocalize with endocytic vesicles containing fluorescently tagged dextran and transferrin in cultured cells (Fig. 6 a). Finally, immunoblot analyses of rat liver membrane fractions for DLP1 and endocytic marker proteins such as α-adaptin, Rab 5, and β-galactosidase (Fig. 6 b), as well as transferrin, Rab4, Rab7, and Rab9 (data not shown), suggest that DLP1 is not enriched in fractions containing these endocytic markers. Dynamin was found predominantly in plasma membrane and Golgi fractions, which is consistent with its proposed participation in the formation of vesicles from the plasma membrane (Kosaka and Ikeda, 1983a ,b; van der Bliek et al., 1993; Damke et al., 1994; Takei et al., 1995) and the Golgi apparatus (Henley and McNiven, 1996; Maier et al., 1996). In support of the immunoblotting data, double immunofluorescence staining of cultured cells with antibodies to DLP1 and clathrin or dynamin was performed (Fig. 7). From these studies we conclude that DLP1 clearly associates with a population of vesicles that is both smaller in size and distinct from those associated with clathrin and dynamin at the plasma membrane.

An ER-associated Dynamin-like Protein

Although DLP1 does not appear to associate with an endocytic compartment, additional data using a variety of different experimental approaches strongly suggest that this dynamin-like protein associates with the ER and small discrete vesicles that associate with microtubules. This conclusion is based on the immunoblotting of subcellular fractions (Fig. 6), as well as the fact that the long strings of DLP1 spots are always associated with the ER tubules (Figs. 9 and 10), whereas individual vesicles are on single microtubules (Fig. 8). It is important to emphasize that the localization of DLP1 at the ER tubules is not coincidental overlap, as supported by the immunoelectron microscopy data described in Fig. 11.

Presently, the identity and composition of the DLP1-positive cytoplasmic vesicles is not clear although biochemical characterization is currently underway. It is interesting that DLP1 and Rab1 are on the same ER cisterna but do not overlap with each other (Fig. 10). In fact, both antigens appear as distinct patches or vesicles that align along the ER in a striking alternating pattern. Whether Rab1 and DLP1 function sequentially in the same process or totally distinct processes is unclear. It is most attractive to speculate that DLP1 and Rab1 act together to mediate the formation and liberation of vesicles from the ER where they could become free to attach to microtubules for subsequent transport to the Golgi complex via pre-Golgi compartments (Presley et al., 1997). Indeed, studies using time-lapse fluorescence microscopy of GFP-DLP1 in living cells demonstrate that these vesicles are motile and move unidirectionally from the cell periphery toward the perinuclear region (results not shown). This observation is consistent with the prediction that DLP1 vesicles are transported from peripheral ER elements to the Golgi compartment via microtubule tracks. Future studies using inhibitory DLP1 antibodies and mutant DLP1 expression constructs will test these predictions and define if this protein participates in anterograde and/or retrograde vesicle flux through the early secretory pathway. Taken together, the biochemical and morphological data presented here raise the exciting possibility that DLP1 is an ER equivalent of the conventional plasma membrane dynamin and participates in the formation of nascent secretory vesicles from the ER cisternae. Functional studies are underway to test this prediction.

The identification and initial characterization of the dynamin-like protein reported here supports the central premise that the dynamin family of proteins may be extensive in number as proposed by McNiven and coworkers (Henley and McNiven, 1996; Urrutia et al., 1997). Based on these findings and the recent proliferation of many different protein families, such as the cytoplasmic myosins, kinesins, adaptins, ARFs, Rabs, syntaxins, and others that are involved in protein trafficking, the number of dynamin family members will likely continue to expand in the near future. Further investigations will define whether DLP1 participates in vesicle scission or other vesicle trafficking processes such as targeting or transport.