INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Dynamin is a 100-kDa GTPase, required for clathrin-mediated endocytosis (De Camilli et al., 1995; Schmid, 1997; Urrutia et al., 1997). Dynamin assembles into a multimeric spiral at the neck of budding vesicles (Takei et al., 1995). Presumably, constriction of the dynamin spiral, driven by GTP hydrolysis, pinches vesicles off from the plasma membrane. This view is supported by a wealth of biochemical, cell culture, and genetic data. The link with endocytosis was made with the discovery that Drosophila shibire defects were caused by mutations in the dynamin gene (Chen et al., 1991; van der Bliek and Meyerowitz, 1991). The shibire mutants are rapidly paralyzed when the pool of synaptic vesicles is depleted by a temperature-sensitive block in recycling via clathrin-mediated endocytosis (Poodry and Edgar, 1979; Kessel et al., 1989; Narita et al., 1989). Mammalian cells transfected with a dominant dynamin mutant are similarly blocked in endocytosis (Herskovits et al., 1993; van der Bliek et al., 1993). Nerve termini incubated with GTP-S show tubular invaginations coated with dynamin spirals, apparently frozen in the act of pinching off (Takei et al., 1995). Purified dynamin also forms spirals and some of these spirals appear partially constricted (Hinshaw and Schmid, 1995). More recently, it was shown that brain cytosol and even purified dynamin alone form vesicles when incubated with exogenous membrane (Sweitzer and Hinshaw, 1998; Takei et al., 1998). Earlier electron micrographs of shibire flies showed electron-dense collars at the necks of budding vesicles (Kosaka and Ikeda, 1983), but their significance was appreciated only after the discovery of dynamin spirals.

We recently described a Caenorhabditis elegans mutant with a defect in dynamin that causes temperature-sensitive paralysis similar to shibire flies (Clark et al., 1997). C. elegans appears to have a single dynamin gene, dyn-1, which is expressed at high levels in the nervous system. Dynamin is also highly abundant in Drosophila and mammalian neurons where it is concentrated at synapses, possibly reflecting the high demand on endocytosis from the recycling of synaptic vesicles (Scaife and Margolis, 1990; McPherson et al., 1994; Estes et al., 1996). For dynamin to function in the synapse, it must be transported from the cell body where it is synthesized along the axonal process to the synapse. Axonal transport could occur through kinesin-dependent mechanisms, which are relatively fast, or through the so-called "slow transport" mechanism, which transports other cytosolic proteins like clathrin (Terada et al., 1996). Once dynamin reaches the synapse, it becomes sequestered in a cytosolic matrix (Estes et al., 1996). From there it can be quickly mobilized to become associated with clathrin-coated pits at the plasma membrane. One could envisage as many as three different localization signals within dynamin: 1) a signal that delivers dynamin to the synapse, 2) a signal that helps sequester dynamin in the synaptic cytosol, or 3) signals that direct dynamin molecules to a specific site on the plasma membrane for assembly into a multimeric complex. Each step could determine where and how much vesicle recycling takes place.

Dynamin has five distinct protein domains that have the potential to contribute to varying degrees to synaptic localization. At the N terminus, the first 300 amino acids make up the GTPase domain, which is highly conserved between dynamin-related proteins, constituting a distinct subgroup within the GTPase superfamily. The second domain, which we call the middle domain, has no known function. The third domain is a pleckstrin homology (PH) domain that binds to inositol phosphates and therefore may be important for interactions between dynamin and the plasma membrane (Salim et al., 1996; Artalejo et al., 1997). The fourth domain is a putative coiled coil that binds to the GTPase and to the middle domain (Smirnova and van der Bliek, unpublished results). Because the putative coiled coil is likely to play a role in forming dynamin multimers, we call this segment the assembly domain. The last domain is a proline-rich domain (PRD)¹, for which coprecipitation experiments showed binding to the Src homology 3 (SH3) domains of amphiphysin (David et al., 1996), Grb2 (Gout et al., 1993), and dynamin associated protein 160 (DAP160) (Roos and Kelly, 1998). C-terminal deletions showed that the PRD is necessary for the localization of dynamin to clathrin-coated pits (Shpetner et al., 1996; Okamoto et al., 1997). However, the strong synaptic localization in neurons suggests that other factors in addition to membrane targeting signals may be equally important in determining the distribution of dynamin.

We set out to identify parts of dynamin that are necessary for synaptic localization in C. elegans with the assumption that targeting to clathrin-coated pits is only one of a series of steps that also includes axonal transport and sequestration in the presynaptic cytosol. Knowing the different targeting mechanisms may help our understanding of synaptic function. In the present study of dynamin localization, we found that dynamin accumulates in the synapse-rich regions of the C. elegans nervous system, as it does in neurons of other organisms. To identify the localization signals contained within dynamin, each of the protein domains was fused to green fluorescent protein (GFP), and their subcellular distribution was determined in single neurons. The degree of localization was quantified with a new application of confocal microscopy in which we compared the fluorescence intensity of a single synaptic patch with the fluorescence intensity of an adjacent segment of the axonal process. The action of several domains of dynamin seems necessary for the protein to be optimally transported from the cell body to the nerve ring. The GTPase domain provided the most potent localization activity, revealing a novel function for this domain.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

C. elegans Strains

Worms were grown on agar plates seeded with Escherichia coli strain OP50 as described (Sulston and Hodgkin, 1988). The wild-type strain was Bristol N2. The dynamin mutant dyn-1(ky51) was described previously (Clark et al., 1997). Mutant strain dpy-20(e1282) was kindly provided by P.W. Sternberg (California Institute of Technology, Pasadena, CA), and unc-104(rh126) was kindly provided by E. Hedgecock (Johns Hopkins University, Baltimore, MD). Other strains were provided by the Caenorhabditis Genetics Center (University of Minnesota, Saint Paul, MN) stock center.

Microinjection Procedures and Expression Constructs

Transgenic worms were obtained by microinjecting 1 ng/µl expression construct together with marker DNA. We used 50 ng/µl plasmid pRF4, which encodes the dominant rol-6(su1006) marker (Mello et al., 1991), or 20 ng/µl plasmid pMH86, which was used to rescue dpy-20(e1282) animals (Han and Sternberg, 1990), and 80 ng/µl pBluescript (Stratagene, La Jolla, CA) as carrier. The pPD series of expression vectors were kindly provided by A. Fire, J. Ahnn, G. Seydoux, and S. Xu (Carnegie Institution of Washington, Baltimore, MD). DNA fragments were recloned by standard procedures. Amplification to fuse DNA fragments or to add restriction enzyme sites was done by PCR with Pyrococcus furiosa DNA polymerase (Pfu) (Stratagene). The new clones were checked by sequence analysis. Boundaries of the fragments used for making the chimeric constructs are shown in Figure 6, and primer sequences are listed in Table 1. Expression was driven by the mec-7 gene promoter (Hamelin et al., 1992) or by the dyn-1 gene promoter. Dynamin protein domains are abbreviated as GTPase, M, A, and PRD. The individual constructs were made as follows.

First Intermediate Plasmid. GFP sequences were amplified from pPD93-65 using primers A and C, cut with BamHI and SphI, and cloned into the dynamin gene construct pCDG1 (Clark et al, 1997), cut with BclI and SphI.

Second Intermediate Plasmid. GFP sequences were amplified from pPD93-65 with primers B and E, and a dynamin gene fragment was amplified from pCDG1 with primers F and G. These two PCR products were fused by reamplification with primers B and G and were cloned into pPD96-41 with NheI and KpnI.

dyn-1::GFP::Dynamin. A 6.2-kb NcoI-KpnI fragment of mec-7::GFP::Dynamin was recloned into the first intermediate plasmid cut with the same enzymes.

dyn-1::GFP::GTPase-M-PH-A. A 3.9-kb NcoI-KpnI fragment from mec-7::GFP::GTPase-M-PH-A was recloned into the first intermediate plasmid cut with the same enzymes.

dyn-1::GTPase::GFP. A 3.9-kb XbaI-BspEI fragment from pCDG1 was ligated to pPD95-67 cut with XbaI and AgeI.

dyn-1::GFP. A 275-bp fragment of GFP was amplified with primers AA and BB, and a 300-bp fragment of dynamin was amplified with primers CC and DD using dyn-1::GFP::Dynamin as template. The two fragments were fused by PCR with primers CC and BB and then cloned into dyn-1::GFP::Dynamin with ClaI and NcoI.

mec-7::GFP. GFP sequences were amplified from pPD93-65 using primers B and D and cloned into pPD96-41 with NheI and KpnI.

mec-7::GFP::Dynamin. A 4.7-kb PacI-KpnI fragment from pCDG1 was cloned into the second intermediate plasmid, cut with the same enzymes.

mec-7::GFP::GTPase-M-PH-A. The assembly domain was amplified from pCDG1 with primers U and V, which introduces a stop codon at the end of the assembly domain. This 295-bp fragment was cloned into mec-7::GFP::Dynamin with HpaI and KpnI.

mec-7::GFP::GTPase. The GTPase was amplified from pCDG1 with primers H and L and then cloned into the second intermediate plasmid with PacI and KpnI.

mec-7::GFP::PH-A-PRD. A 3.4-kb BclI-KpnI fragment from pCDG1 was recloned into BamHI-KpnI-cut mec-7::GFP. To correct the reading frame between GFP and dynamin sequences, part of GFP and the linker sequence were reamplified with primers J and K and then recloned with NcoI and BclI.

mec-7::GFP::M. The middle domain was amplified from pCDG1 with primers M and I and then cut with BclI and ligated into BamHI-cut mec-7::GFP. A stop codon was introduced by ligating primer T into the KpnI site.

mec-7::GFP::PH. The PH domain was amplified from pCDG1 with primers N and O and then cut with BclI and ligated to BamHI-cut mec-7::GFP.

mec-7::GFP::A. The assembly domain was amplified from pCDG1 using primers P and Q and then cut with BclI and ligated to BamHI-cut mec-7::GFP.

mec-7::GFP::PRD. A 69-bp sequence was amplified from pCDG1 using primers R and S and then cut with BclI and AgeI and ligated into dyn-1::GFP, which had been cut with BamHI and AgeI, to make dyn-1::GFP::PRD. This plasmid contains a 2.8-kb NcoI-KpnI fragment, which was recloned into mec-7::GFP cut with the same enzymes.

mec-7::GFP::GTPase(K46A). A 1.1-kb fragment was amplified from dyn-1::GFP::Dynamin using primers A and EE. A 520-bp fragment was amplified from the same template, but with primers G and FF. The two fragments were fused by amplification with primers A and G. The fusion product was cut with NcoI and PacI and ligated into mec-7::GFP::GTPase cut with the same enzymes.

mec-7::GFP::M-PH-A-PRD. A 1.5-kb NcoI-HindIII fragment from mec-7::GFP::M was cloned into the first intermediate plasmid cut with the same enzymes to give dyn-1::GFP::M. A 3.8-kb BspE1-HindIII fragment from dyn-1::GFP::Dynamin was cloned into dyn-1::GFP::M cut with the same enzymes. This construct was then cut with NcoI and KpnI, and the resulting 4.9-kb fragment was ligated into NcoI-KpnI-cut dyn-1::GFP to make dyn-1::GFP::M-PH-A-PRD. Finally, a 2.4-kb fragment was cut out of dyn-1::GFP::M-PH-A-PRD with NcoI and HpaI and ligated into mec-7::GFP::PH-A-PRD cut with the same enzymes.

Indirect Immunofluorescence

To generate anti-dynamin antibodies, we expressed the C-terminal half of C. elegans dynamin in E. coli with the bacterial expression vector pQE30 (Qiagen, Valencia, CA). This vector adds six histidines, which we used to purify the recombinant protein by Ni-affinity chromatography. Rabbit antisera were generated by Cocalico Biologicals (Reamstown, PA) and then blot purified with a dynamin protein fragment. Anti-synaptotagmin antibodies were kindly provided by Mike Nonet (Washington University, St. Louis, MO). Secondary antibodies (Boehringer Mannheim, Indianapolis, IN) were preadsorbed with acetone-powdered C. elegans to remove cross-reacting antibodies (Miller and Shakes, 1995).

Immunofluorescence procedures, adapted from Finney and Ruvkun (1990), were as follows. Well-fed worms were washed by pelleting and resuspending in water and then permeabilized by three cycles of freezing and thawing in 10 ml fixative (0.2 M Na-K-phosphate, pH 7.2, 4% paraformaldehyde). After three washes in 10 ml 100 mM Tris/Cl (pH 7.5), 1 mM EDTA, 1% Triton X-100 (buffer I), the worms were resuspended in 750 µl buffer I and broken with six strokes of a Dounce homogenizer (Corning Glass, Corning, NY). The worms were then incubated for 2 h at 37°C in 10 ml 0.1 M Tris/Cl (pH 6.9), 5% -mercaptoethanol, and 1% Triton X-100, washed three times in 10 ml buffer I, followed by 2 h at 25°C in 10 ml 10 mM DTT and 1 M borate, and again washed three times in 10 ml buffer I. The worms were then gently agitated for 1-2 h at 37°C in 3 ml 10 mg/ml collagenase, 0.1 M Tris/Cl (pH 7.5), and 1 mM CaCl₂, followed by three washes in PBS, 3 h at 0°C in 10 ml fixative with 10 mM EGTA, and three more washes in PBS. The worms were then incubated for 16 h at 25°C in 300 µl PBS with 1% BSA, 0.5% Triton X-100, and 0.05% NaN₃ with 3 µl primary antibody, followed by three washes in 5 ml PBS with 0.2% BSA, 0.5% Triton X-100, and 0.05% NaN₃ and incubated for 3 h at 37°C with secondary antibodies in PBS with 1% BSA, 0.5% Triton X-100, and 0.05% NaN₃. After three washes in 10 ml buffer I and one wash in 1 ml mounting buffer (Molecular Probes, Eugene, OR), the worms were resuspended in Antifade (Molecular Probes) and mounted on slides coated with a thin film of dried agarose.

Microscopy and Image Analysis

Immunofluorescence was observed with a Nikon (Garden City, NY) FXA microscope equipped with filters for rhodamine and FITC. GFP was observed with an FITC excitation filter and a wide-band emission filter (Nikon B2A) so that GFP could be distinguished from the orange-tinted autofluorescence of gut granules. Confocal images were collected with a Zeiss LSM 310 microscope (Carl Zeiss, Thornwood, NY) in series of 0.75-µm optical sections that were combined into one image with the LSM software. The average intensities within a circled area covering a synaptic patch and a boxed area of identical size along the axonal process were measured with NIH Image software. The relative fluorescence intensities were then determined with a calibration plot made by imaging a series of fluorescent beads (Microscope Image Intensity Calibration Kit, Molecular Probes) using the same contrast and intensity settings of the confocal microscope as were used for the original image. Where indicated, small aggregates of GFP were eliminated by incubating the animals for 24 h at 20°C in M9 medium with 5% DMSO. A slurry of freshly grown bacteria (E. coli strain OP50) was added as food.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Distribution of Dynamin in C. elegans Determined by Immunofluorescence

We previously showed that dynamin is expressed at high levels in the C. elegans nervous system using the dynamin gene promoter fused to -galactosidase (Clark et al., 1997). Here, we used immunofluorescence with an anti-dynamin antibody to investigate the subcellular distribution. We detected high levels in the nerve ring, along the ventral nerve cord, the dorsal nerve cord, and in pharyngeal neurons (Figure 1). The C. elegans nerve ring is a large ganglion encircling the pharynx and consists largely of axonal processes with their many synapses (White et al., 1986). The nerve ring is devoid of cell bodies. Many of these cell bodies are in the head but clearly separated from the nerve ring. The concentration of fluorescence in the nerve ring indicated that dynamin was highly localized to synapse-rich regions. In some preparations, we also detected regularly spaced patches of immunofluorescence along sublateral neurons in the head (Figure 1). These patches are consistent with the location of chemical synapses detected by electron microscopy and by immunofluorescence of other synaptic proteins (Hall and Rand, personal communication; Nonet et al., 1997). The localization of dynamin to chemical synapses is similar to the localization of other presynaptic proteins such as synaptotagmin (Nonet et al., 1993).

View larger version (79K): [in this window] [in a new window]   Figure 1.   Endogenous dynamin detected in C. elegans by immunofluorescence. A wild-type (N2) animal was stained with anti-dynamin antibodies. The immunofluorescence is concentrated in the nerve ring (nr) and along ventral (vc) and dorsal (dc) nerve cords. The punctate pattern in the pharynx corresponds to pharyngeal neurons (pn). Dynamin was also detected in sublateral neurons (sn). The inset shows that dynamin accumulates along the apical surface of intestinal cells (arrowhead). In this case, the worm was broken during the permeabilization procedure, which helped unmask the specific intestinal staining.

Non-neuronal expression was difficult to ascertain by immunofluorescence with anti-dynamin antibodies, although our previous experiments with the dyn-1 promoter fused to -galactosidase also showed expression in non-neuronal cell types. In a few preparations, staining was observed along the apical surface of intestinal cells (Figure 2, inset), but more typically this staining was obscured by autofluorescence, which was also detected in control experiments omitting the primary antibody or blocking with recombinant dynamin protein (our unpublished results). Autofluorescence is largely due to the accumulation of lipofuscin in secondary lysosomes of gut granules (Clokey and Jacobson, 1986). Despite this technical difficulty, it seems likely that the dynamin gene is ubiquitously expressed, because dynamin is required for all clathrin-mediated endocytosis. As described in the next section, a more comprehensive description of dynamin localization was obtained with the dyn-1 GFP fusions. However, the immunofluorescence results do establish the subcellular localization of endogenous dynamin in neurons, which was necessary to ensure the validity of subsequent localization experiments using GFP-chimeras.

View larger version (28K): [in this window] [in a new window]   Figure 2.   Dynamin localization unaffected in an unc-104 worm. An unc-104 mutant animal was double labeled with anti-synaptotagmin (A) and anti-dynamin antibody (B). The anti-synaptotagmin antibody showed diffuse immunofluorescence in the head and concentrated immunofluorescence in the cell bodies of the ventral nerve cord (vc), which reflects the mislocalization of synaptic vesicles in unc-104 mutants. The anti-dynamin antibody showed immunofluorescence concentrated in the nerve ring (nr) and evenly distributed immunofluorescence along the ventral nerve cord (vc), similar to wild-type localization. The staining in the gut is mostly nonspecific, because it can also be detected in control experiments leaving out primary antibody (our unpublished results). In wild-type worms, which are shown here as a control, anti-synaptotagmin and anti-dynamin staining are both concentrated in the nerve ring (insets).

To determine whether dynamin uses the same axonal transport mechanism as synaptic vesicles, we investigated the dynamin distribution in unc-104 mutant animals in which synaptic vesicles stay in the neuronal cell bodies instead of being transported out to the synapses (Hall and Hedgecock, 1991). The unc-104 gene encodes a kinesin-like protein required for axonal transport of synaptic vesicles. Synaptic vesicles can be detected by immunofluorescence with antibodies directed against synaptotagmin (Nonet et al., 1993). In wild-type worms, the immunofluorescence with anti-synaptotagmin antibody is concentrated in the nerve ring and along the nerve cords in a pattern similar to that of dynamin (Figure 2, A and B, insets). In unc-104 mutants, synaptotagmin was mislocalized to cell bodies, which were detected as fluorescent spots throughout the head (Nonet et al., 1993). Synaptotagmin immunofluorescence was also concentrated in spots corresponding to neuronal cell bodies along the ventral nerve cord of unc-104 mutant animals (Figure 2A).

In contrast to synaptotagmin, the distribution of dynamin was unaltered in unc-104 animals, showing fluorescence concentrated in the nerve ring and evenly distributed along the ventral nerve cord (Figure 2B). This suggests that dynamin is not transported by the unc-104 kinesin, but instead uses some other mechanism. One such transport mechanism is the so-called slow transport mechanism, which might be important for dynamin localization, because it is also used by other cytosolic proteins, such as clathrin (Terada et al., 1996).

Distribution of the GFP-Dynamin Chimera in Neurons and Non-Neuronal Cells

To observe dynamin localization in vivo, we inserted GFP coding sequences between the dyn-1 gene promoter and the dynamin protein coding sequences. Transgenic worms expressing the chimeric protein showed intense green fluorescence in the nerve ring and nerve cords in a pattern similar to that observed by immunofluorescence (Figure 3A). This pattern indicates that the chimeric protein is efficiently transported and perhaps sequestered at the synapse. The GFP-dynamin chimera enabled the detection of dynamin gene expression in non-neuronal cell types that went undetected by immunofluorescence. Autofluorescence, caused by the accumulation of lipofuscin in gut granules (Clokey and Jacobson, 1986), could be distinguished from GFP, because it has a yellow or orange tint when viewed with a broad-pass emission filter. GFP expressed in intestinal cells was made visible by the accumulation of green fluorescence at the apical surfaces (Figure 3B). This accumulation suggests a high rate of endocytosis from the intestinal lumen. We also detected dynamin along the outer membranes of the pharynx (Figure 3A), the gonadal sheath cells (Figure 3C), the spermathecae (Figure 3C), and in coelomocytes, which are scavenger cells in the C. elegans body cavity (Figure 3D). The expression in male animals was similar to that in hermaphrodites in their nonreproductive tissues (our unpublished results). Males also expressed GFP in cells lining the seminal vesicle and the vas deferens, and GFP forms aggregates at the point where spermatocytes bud to become spermatids before entering the seminal vesicle (our unpublished results). It is possible that these aggregates are part of the residual cytoplasmic body, which is left behind after spermatids bud from the rachis (Ward et al., 1981), even though transgenes are usually not expressed in the germ line (Kelly et al., 1997).

View larger version (37K): [in this window] [in a new window]   Figure 3.   Localization of a GFP-dynamin chimera under control of the dyn-1 promoter. (A) Close-up of the head region, showing that GFP-dynamin is concentrated in the nerve ring (nr). GFP-dynamin is also detected along the outside of the pharynx (ph). The yellow or orange staining of the gut is largely due to autofluorescence of gut granules. (B) Close-up of a midsection of the body, showing that the fluorescence of GFP-dynamin appears punctate along the ventral nerve cord (vc) and along the dorsal nerve cord (dc). GFP is also detectable along apical surface of the intestinal cells (ic). Some of the intestinal GFP may have been masked by the autofluorescence of gut granules (yellow-orange staining). (C) Close-up of a midsection of the body, showing fluorescence of GFP-dynamin in a spermatheca (sp) and along the gonadal sheath (gs). (D) Close-up of a midsection of the body, showing fluorescence of GFP-dynamin in a coelomocyte (co).

The amount of GFP-dynamin chimera as determined by Western blotting was typically between 0.2 and 0.8 times the amount of endogenous dynamin (our unpublished results). This level did not alter the temperature-sensitive paralysis of the dyn-1 (ky51) allele, nor did it rescue the embryonic lethal phenotype of a null allele isolated in our lab (our unpublished results). We conclude that GFP does not cause mislocalization or dominant interference, but it does interfere with the endocytic function of the attached dynamin. Similar results were obtained with GFP fused to phragmoplastin (a dynamin-like protein) in transgenic plants (Gu and Verma, 1997).

To identify parts of dynamin that conferred localization, we tested chimeric constructs in which portions of the dynamin sequence were deleted. When we tested the GFP fused to the GTPase domain, we found that this was sufficient for correct localization in neurons and intestinal cells (Figure 4C). The pattern of autofluorescence, the pattern obtained with full-length dynamin and that obtained with GFP alone are shown for comparison in Figure 4, A, B, and D. The amount GFP-GTPase localized to the nerve ring was comparable with the amounts in the surrounding neuronal cell bodies. In intestinal cells, however, localization of the GTPase domain was much more striking, showing strong fluorescence along the apical brush border. This result indicates that the GTPase domain is important for the localization of dynamin.

View larger version (42K): [in this window] [in a new window]   Figure 4.   Localization of GFP chimeras expressed with the dynamin promoter. The left panels show the heads of transgenic animals with arrows pointing at the nerve rings. The right panels show midsections of transgenic animals with arrowheads pointing at the intestinal lining. (A) Worm with no GFP. The yellow or orange spots are gut granules, which are autofluorescent. (B) Worm containing GFP fused to full-length dynamin. The nerve ring is strongly fluorescent and there is some fluorescence detectable along the apical lining of intestinal cells (arrowheads). (C) Worm containing GFP fused to the GTPase domain. The nerve ring is still detectable, but additional diffuse fluorescence throughout the head suggests a certain degree of mislocalization. Localization to the intestinal lining was evident (arrowheads). (D) Worm containing GFP alone expressed under control of the dyn-1 promoter. There was no detectable nerve ring fluorescence, nor was GFP detectable along the intestinal lining, although it clearly was expressed in neurons and gut cells.

Subcellular Localization of Dynamin in ALM Neurons

Localization of specific parts of dynamin could occur if the domain in question contains a specific targeting signal or by association with endogenous dynamin. The latter possibility needed further consideration, because it was known that dynamin forms a multimeric complex (Tuma and Collins, 1994; Hinshaw and Schmid, 1995). Because more than one domain could participate in multimerization and targeting, it was necessary to determine the contributions of each individual dynamin domain separately. To obtain accurate information about the contributions of different domains of dynamin to synaptic localization, we generated a series of chimeras with the mec-7 promoter fused to GFP and to the individual dynamin domains. Because the activity of the mec-7 promoter is restricted to six touch cells (Hamelin et al., 1992; Chalfie et al., 1994), the promoter fusions allowed us to focus on a single pair of easily identifiable neurons, the ALMs, which have their cell bodies located just anterior of the vulva (White et al., 1986). Each ALM neuron sends a process anteriorly along the lateral nerve cord ending close to the tip of the nose. A single branch enters the nerve ring and curves ventrally where it meets the AVM neuron. Electron microscopic analysis has shown presynaptic varicosities corresponding to three clusters of chemical synapses in the branches of the ALM neurons (White et al., 1986).

GFP fused to full-length dynamin under the control of the mec-7 promoter gave strong fluorescence in selected patches along the branches of the ALM neurons (Figure 5B). Similar patches were observed with GFP fused to the synaptic vesicle protein VAMP/synaptobrevin (Nonet et al., 1998), although there was not enough fluorescence for quantitation (our unpublished results). These patches are likely the chemical synapses of the ALM neurons, because their size, number, and location were consistent with those detected by electron microscopy (White et al., 1986). With some expression constructs, we also saw fluorescence in a few large spots in the cell body or along the axonal process in numbers that varied between animals and in locations that were clearly separated from synapses. These spots may correspond to protein aggregates, autophagosomes (Hollenbeck, 1993), or possibly axonal "traffic jams" as described in a Drosophila kinesin mutant (Hurd and Saxton, 1996). With one chimera (mec-7::GFP::GTPase) we detected punctate fluorescence throughout the ALM neurons. Exposing the worms to DMSO reduced the number of spots, which suggests that the spots were protein aggregates (our unpublished results). Fortunately, there was no indication that the spots affected the specific localization of our GFP-chimeras to the synapses.

View larger version (35K): [in this window] [in a new window]   Figure 5.   Green fluorescence concentrated in presynaptic varicosities of ALM neurons. (A) Distribution of GFP near the distal end of an ALM neuron. This is a close-up of the nerve ring of a transgenic worm containing the mec-7::GFP construct. A segment of the ALM axonal process is shown at the top. The branch, which enters the nerve ring, is shown making a curve toward the bottom. The synaptic clusters are visible as three patches of fluorescence along the branch. The image is a composite in which a stack of confocal sections was merged to visualize the curved branch of the ALM process. (B) Distribution of a GFP-dynamin chimera in part of anALM neuron. This is a close-up of the nerve ring of a transgenic worm containing the mec-7::GFP::Dynamin construct. Almost all fluorescence is concentrated in the three patches corresponding to clusters of synapses. (C) Diagram of an ALM process, with a circle around one of the clusters of synapses and a box along the axonal process depicting the areas chosen to quantitate the degree of localization. The measurements were all conducted with the same surface area (200 pixels of images taken with the same magnification).

We used confocal microscopy to quantify the degree of synaptic localization. A three-dimensional representation of the ALM neurons that were expressing the GFP chimeras was made with a series of confocal images. This series of images was converted to a single two-dimensional image, and the fluorescence intensity was determined in two selected areas, one in a synaptic patch and one in an adjacent part of the axonal process (Figure 5C). The occasional aggregates that were visible as fluorescent spots along the axonal process were avoided, because they might skew the outcome. A calibration curve, made with fluorescent beads, was used to account for the nonlinear relation between pixel values and fluorescence intensity. The degree of localization was expressed as a fluorescence ratio in which the amount of fluorescence in a synapse was divided by the amount of fluorescence in the adjacent axonal process. This approach made it possible to quantify the degree of localization in a highly reproducible manner.

GFP alone does not accumulate in the patches, which correspond to synapses, but is instead distributed in a gradient emanating from the cell body (Figure 5A). The fluorescence intensity in the synaptic patches was very close to that in the axonal process, giving an average fluorescence ratio of 1 (Figure 6B). In marked contrast to the uniform distribution of GFP by itself, the fusion between GFP and dynamin was 17 times more concentrated in synapses than in adjacent sections of the axonal process (Figures 5B and 6B). This demonstrated that axonal transport and synaptic sequestration were not saturated by ectopic expression with the mec-7 promoter and conversely that these mechanisms were able to localize the GFP-dynamin chimera in ALM neurons.

View larger version (34K): [in this window] [in a new window]   Figure 6.   Localization of GFP-dynamin chimeras in ALM neurons. (A) Diagram of chimeric proteins used to study localization in ALM neurons. GFP was fused to different portions of dynamin as indicated by the amino acid numbers. The linker sequences, shown with the one-letter amino acid code, were inserted with oligonucleotides. All constructs were driven by the mec-7 promotor. (B) Histogram showing the localization of GFP fusion proteins in ALM presynaptic varicosities relative to an adjacent section of the axonal process as indicated in Figure 5. These values were averages of 6-16 worms and are shown with the SE.

We tested the contribution of the individual dynamin protein domains by analyzing the distribution of GFP chimeras in ALM neurons (Figure 6). Localization, expressed as fluorescence in the synapse relative to fluorescence in the process, varied from onefold with the PH domain or PRD to sevenfold with the GTPase domain (Figure 6B). Although Western blotting verified that all chimeras were intact (our unpublished results), we could not rule out the possibility that the PH domain and PRD were misfolded or otherwise impaired by GFP. The lack of synaptic localization of these two constructs, of GFP alone, and of a GFP--galactosidase chimera (our unpublished results) provides a compelling argument that the localization caused by the other domains is due to a specific concentrating process. We conclude that three domains, the GTPase, the assembly, and to a lesser degree the middle, were each sufficient for specific localization to the synaptic clusters (seven-, four-, and twofold, respectively), whereas the PH domain and PRD were not (Figure 6B).

Localization of the GTPase domain in ALM neurons is consistent with the localization of chimeras expressed by the dyn-1 promoter (Figure 4C). To test whether localization was GTP dependent, we introduced the K46A mutation, which presumably prevents GTP binding by affecting the G1 consensus motif. This mutation was previously shown to block dynamin function, but not assembly into a multimeric spiral (van der Bliek et al., 1993; Warnock et al., 1996). The localization factor was reduced from sevenfold for the wild-type GTPase to fourfold for the K46A mutant, which suggests that GTP binding does influence localization but is not the only determinant (Figure 6B).

The finding that individual domains of dynamin were not localized to the same extent as full-length dynamin suggests that different domains act synergistically or additively, depending on whether they participate in the same process or in sequential transport events. We found that the individual localization factors were not additive when different domains were combined (Figure 6). We also found that the localization factor was influenced by the position of GFP (our unpublished results). We therefore focused on constructs containing GFP fused to the N termini of different parts of dynamin. A chimera with all but the PRD (mec-7::GFP::GTPase-M-PH-A) was still 11-fold more concentrated in synaptic clusters than in the axonal process (Figure 6B), consistent with nerve ring localization that could be observed with the dyn-1 promoter (our unpublished results). Deleting the GTPase domain in mec-7::GFP::M-PH-A-PRD decreases the localization factor from 17-fold to 2.5-fold as expected if the GTPase domain contains a localization signal (Figure 6B).

Our findings suggest complex synergy in the localization of full-length dynamin, for example, if interactions between multiple domains were required for assembly into a multimeric complex. We conclude that three of the five domains by themselves were sufficient for localization, but that the combined action of multiple domains was necessary for maximal localization.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Neuronal Localization of Dynamin

Our experiments explored the subcellular distribution of dynamin in C. elegans and its underlying causes. First, the immunofluorescence and GFP chimeras showed that dynamin is concentrated in parts of the nervous system that are rich in chemical synapses. Second, it was possible to test the contributions of the individual protein domains to localization in ALM neurons. No fewer than three of the five dynamin protein domains contribute to localization as determined by fluorescence intensity. The GTPase domain showed the highest degree of synaptic localization and was also specifically localized along the apical surface of intestinal cells. This was unexpected, because in previous experiments with transfected mammalian cells, deleting the PRD abolished localization to coated pits, and further deletions caused dynamin to lose all membrane association (Shpetner et al., 1996; Okamoto et al., 1997). However, our experiments did not address membrane localization, but rather localization to specialized parts of the cell, such as the presynaptic cytosol and the apical lining of intestinal cells. Therefore, our results were complementary to the results obtained by transfecting dynamin deletions into fibroblasts. Our discovery that the GTPase domain confers strong localization in situ in C. elegans tissue was most revealing, because it may lead to new factors that contribute to the localization process.

The strong immunofluorescence in the nerve ring and along the nerve cords most likely reflects the localization of dynamin to neuronal synapses (Figure 1). This is particularly clear for the nerve ring, which is largely devoid of cell bodies and instead consists primarily of processes connected by chemical and electrical synapses (White et al., 1986). The distribution of the GFP-dynamin chimera, as seen in detail in touch cells, is also consistent with presynaptic localization, because it matches that of synaptic vesicles detected by electron microscopy (White et al., 1986) and by VAMP-GFP. The distribution of C. elegans dynamin is similar to that in mammals and Drosophila, in which dynamin is highly concentrated in presynaptic cytosol, consistent with the important role that dynamin plays in synaptic vesicle recycling (Scaife and Margolis, 1990; McPherson et al., 1994; Estes et al., 1996). This distribution raises the question of how soluble proteins such as dynamin are transported to and become sequestered in the presynaptic cytosol.

Our analysis of unc-104 animals indicates that dynamin is not transported together with synaptic vesicles, because synaptotagmin was clearly mislocalized, whereas dynamin was not (Figure 2). It remains possible that other kinesins transport dynamin, or that the protein is sequestered in the presynaptic varicosities following passive diffusion. However, it seems more likely that dynamin uses slow axonal transport, because the bulk of soluble proteins such as clathrin and synapsin I follow this route (Terada et al., 1996).

Distribution in Non-Neuronal Cells

GFP-dynamin under control of the dyn-1 promoter showed expression in many non-neuronal cell types (Figure 3). GFP proved to be more sensitive than immunofluorescence, because GFP-expressing worms had less background fluorescence and were not subjected to harsh permeabilization procedures. Nevertheless, the expression patterns observed with immunofluorescence and GFP both agree with previous -galactosidase staining, showing high levels in neurons and lower levels in other cell types (Clark et al., 1997). Most likely the dyn-1 gene is ubiquitously expressed, because dynamin is essential for all clathrin-mediated endocytosis (Herskovits et al., 1993; van der Bliek et al., 1993), and we know of only one dynamin gene in C. elegans (Clark et al., 1997). The dyn-1 gene is most likely nonredundant, because a dyn-1 null allele, recently isolated in our laboratory, is embryonic lethal (our unpublished results). We detected expression in many of the same cells that were detected previously with -galactosidase staining, including pharyngeal muscles and intestinal cells (Clark et al., 1997). However, we also detected expression in coelomocytes, spermathecae, and gonadal sheath cells. These may have been missed with -galactosidase staining, because this procedure exhibits a threshold effect that exaggerates differences in expression levels. More importantly, the GFP-dynamin experiments also showed much more distinct subcellular localization than seen with immunofluorescence.

GFP-dynamin had a punctate distribution in coelomocytes, which might correspond to clathrin-coated pits (Figure 3D). Coelomocytes contain many coated pits, which are used to scavenge the pseudocoelomic cavity (White, 1988). A punctate distribution was also detected in spermathecae and pharyngeal muscles, where GFP-dynamin is localized to the surface facing the body cavity (Figure 3, A and C). However, it is unclear why spermathecae and pharyngeal muscles would have high rates of endocytosis. It is much easier to understand why intestinal cells have high levels of dynamin (Figure 3B). Here, GFP-dynamin was concentrated along the apical surface facing the intestinal lumen, consistent with apical microvilli supporting high rates of endocytosis to retrieve nutrients from the intestinal lumen.

Localization to the apical surface of intestinal cells was even more pronounced in transgenic animals expressing the GTPase-GFP chimera (Figure 4C). The apical lining consists of a brush border, raising the alternative possibility that the GTPase-GFP chimera is bound to a matrix component adjacent to the apical membrane, rather than binding to the membrane itself. Such sequestration may help form a pool of dynamin molecules, held in reserve to support bursts of endocytosis, similar to the pool of dynamin molecules sequestered to a cytosolic matrix component in Drosophila neuromuscular junctions (Estes et al., 1996). Although we could not rule out localization strictly with the plasma membrane, it will be very interesting to determine whether such alternative mechanisms exist outside the nervous system.

Localization of Individual Protein Domains

Fluorescence of the nerve ring and the intestinal lining suggests that the GTPase domain is sufficient for localization (Figure 4C). We envisage three factors that may be important for synaptic localization. First, localization might be the passive consequence of association with dynamin encoded by the endogenous dyn-1 gene. Endogenous dynamin had to be present in all our experiments, because dynamin is essential for cell survival. Second, localization might reflect association with the axonal transport machinery. This mechanism is unlikely to occur in intestinal cells, which do localize the GTPase domain but presumably lack an intestinal equivalent of axonal transport. Third, localization might occur through passive diffusion along the axonal process followed by sequestration, either by a cytosolic matrix component or at the plasma membrane. Thus, different mechanisms may contribute to localization, depending on the specific functions of each individual domain.

Neither the PH domain nor PRD conferred synaptic localization to GFP (Figure 6B). This result was unexpected, because earlier deletion studies with mammalian cells had shown that the PRD is required to localize dynamin to coated pits (Shpetner et al., 1996; Okamoto et al., 1997). Coated pits contain proteins such as amphiphysin and DAP160 that bind to the dynamin PRD through their SH3 domains (David et al., 1996; Roos and Kelly, 1998). These proteins may help direct dynamin to the necks of budding vesicles or control the constriction process in some other way. The PH domain also binds to a membrane component (phosphatidyl inositol 4,5-diphosphate), which may act in concert with the PRD in the final stages of localizing dynamin to coated pits (Barylko et al., 1998). However, our results suggest that the interactions with the PH domain and PRD are not strong enough to sequester the chimeras in presynaptic varicosities. Evidently, other domains, such as the GTPase, middle, and assembly, contribute to synaptic localization.

Yeast two-hybrid and in vitro binding experiments with isolated dynamin fragments show three interactions between different parts of dynamin: the assembly domain binds to itself and to the GTPase and middle domains (Smirnova and van der Bliek, unpublished results). This raises the possibility that these three domains associate with endogenous dynamin and thereby piggyback to presynaptic varicosities. Such a localization mechanism seems likely for the middle and assembly domains, because these two domains showed strong binding. However, binding between the GTPase and assembly domains is relatively weak. Furthermore, the same mutation that decreases the specific localization of the GTPase domain in ALM neurons (mec-7::GFP::GTPase(K46A); Figure 6B) has the opposite effect in the yeast two-hybrid system and in vitro binding experiments. The mutant GTPase domain binds more strongly to the assembly domain (our unpublished results) and was previously shown to stabilize a coassembled dynamin complex (Warnock et al., 1996). This makes it unlikely that the strong localization of the wild-type GTPase domain is solely due to association with endogenous dynamin. An alternative mechanism, such as binding to a cytosolic matrix component, might contribute to the localization of the GTPase domain in neurons and intestinal cells.

Any localization signal that might be contained by the GTPase domain, and possibly by the middle and assembly domains, must be functional both in neurons and in intestinal cells. Most other GTPases, such as ras, do not contain intrinsic localization signals, but members of the rab family of small GTPases are an exception (Novick and Zerial, 1997). Each of these proteins is targeted to a specific membranous compartment by a hypervariable sequence at its C terminus (Chavrier et al., 1991). For example, rab3A and rab3B are targeted to presynaptic vesicles and apical membranes of polarized epithelial cells (Weber et al., 1994), which superficially resembles the targeting of the dynamin GTPase domain that we describe here. However, it seems more likely that the dynamin PH domain and PRD are responsible for the association with coated-pit constituents (phosphatidyl inositol 4,5-diphosphate and SH3 domains), whereas the GTPase domain and perhaps also some of the other dynamin domains provide a novel localization functions. Our results suggest that these localization functions are important in cells with high rates of endocytosis. The sequestration of a large pool of dynamin near the site of endocytosis enables neurons to rapidly regenerate synaptic vesicles in response to increased synaptic activity, whereas intestinal cells may also require localized dynamin to sustain high rates of endocytosis when food becomes available. Distinguishing the contributions of different dynamin domains will help unravel the localization process.