ARTICLE

Early Insights from the Drosophila melanogaster Mutant, shibire^ts (shi^ts)

shi^ts flies were initially isolated as temperature-sensitive paralytic mutants (1). The temperature-dependent paralysis that occurs in these flies results from a depletion of neurotransmitter-containing vesicles at nerve terminals (2). Ultrastructural studies have revealed that this reduction in the number of synaptic vesicles (SVs) is caused by an impairment in vesicle membrane recycling following exocytosis, resulting in an accumulation of "collared pits" at the presynaptic plasma membrane (refs. 3-5; Fig. 1 a-c). Interestingly, these invaginations accumulate in a number of other cell types in these flies, indicating a general defect in endocytosis (refs. 5 and 6; Fig. 1d).

Shortly after the initial purification of dynamin from bovine brain, a cDNA encoding the full rat brain dynamin was isolated by Obar et al. (7) using an anti-dynamin polyclonal antibody. Subsequently, sequence analysis of the D. melanogaster shibire gene revealed an 81% similarity to the brain dynamin cDNA (8, 9). The finding that dynamin is the homologue of the shibire gene product was a significant step toward understanding the functional role of these proteins. Important information on the role of dynamin as an endocytic GTPase can be inferred from the sequence analyses of the two shibire alleles, shi^ts1 and shi^ts2, which occur within or near the GTP-binding motif and render the protein nonfunctional at the restrictive temperature (8). Similar mutations in dynamin overexpressed in mammalian cells inhibit receptor-mediated endocytosis and produce the shi^ts phenotype (10-13). Thus, the extensive degree of structural similarity between dynamin and the shibire gene product suggests that they share functional homology and, as discussed below, substantiates recent experiments testing the involvement of dynamin in receptor-mediated endocytosis.

Identification of Novel Dynamin-Encoding Genes

Because dynamin and its encoding cDNA were both originally isolated from mammalian brain, it was thought to be a neuron-specific protein. Early studies using Western and Northern blot analysis appeared to confirm this prediction (14). Subsequent studies using antibodies and cDNA probes to the enzymatic head domain of dynamin suggested that a highly related protein may be ubiquitous in distribution (15). Therefore, several laboratories initiated a search for additional dynamin-encoding genes that could be expressed in other tissues. Our laboratory (15) and Sontag et al. (16) independently isolated and characterized a novel dynamin-encoding cDNA that is ubiquitously expressed in all rat tissues examined. Sequence analyses revealed that this gene encodes for a protein of 868 aa with a calculated molecular mass of 98.2 kDa and is 79% identical at the protein level to the previously described rat brain dynamin. At the same time, Nakata et al. (17) isolated an additional dynamin cDNA from rat testis. The complete open reading frame encoded for a novel protein of 848 aa, which is approximately 80% identical to the two other dynamin isoforms. Northern blot analysis and in situ hybridization experiments had showed that this gene is expressed exclusively in Sertoli cells of the testis, although a recent study by Cook et al. (18) has demonstrated that this gene is also expressed in the brain and the lung. The degree of identity between the rat brain-specific dynamin and the other dynamin family members is exceptionally high at the amino terminus, which contains the tripartite GTP-binding motif (Fig. 2) and is highly conserved with the shibire gene product. In addition, dynamins also contain a pleckstrin homology motif and a proline-rich carboxyl-terminal tail. To distinguish among the different dynamin isoforms, we will follow the nomenclature proposed by Sontag et al. (ref. 16; Fig. 2) in which the neuron-specific dynamin is called dynamin I (DynI) and the ubiquitously expressed isoform is called dynamin II (DynII). Therefore, the dynamin expressed exclusively in the testis, lung, and brain is referred to as dynamin III (DynIII). Subsequently, additional dynamin isoforms that are yet to be identified would be termed DynIV and so on. Sontag et al. (16) have isolated four different alternatively spliced cDNAs encoding for both DynI and DynII. Cook et al. (18) have indicated that DynIII is also expressed as four different spliced forms. Thus, it is likely that there are at least 12 different dynamin isoforms present in the adult rat brain. The above nomenclature proposes that different splice sites be identified by letters in an amino- to carboxyl-terminal order (e.g., DynIab corresponds to a protein that contains two insertions located at the amino- and the carboxyl-terminal regions; ref. 16).

Multiple Dynamins for Distinct Endocytic Pathways

As described above, there are at least three different dynamin genes with several alternatively spliced forms (Fig. 2). This poses the question as to why are there so many dynamins and whether they perform unique or redundant cellular functions. Because DynII is expressed in all cells and tissues that have been examined (15, 16), it is attractive to predict that it participates in a ubiquitous process such as clathrin-mediated endocytosis (CME). At present, the evidence implicating a dynamin in CME is extensive. Morphological studies of epithelial tissues from shi^ts flies show an accumulation of clathrin-coated pits (refs. 5 and 6; Fig. 1d). This observation is complemented by recent studies in cultured mammalian cells transfected with various dynamin mutations (10-13). Although these later experiments transfected neuronal DynI into epithelial cells, the effects on CME and cellular morphology were dramatic and convincing. Endocytic uptake of labeled transferrin in these cells was markedly reduced, and long tubular endocytic structures (Fig. 3a), reminiscent of those seen in shi^ts flies (Fig. 1), were present. In many instances, clathrin cages appeared to associate with the cytoplasmic ends of these membranous invaginations, which were labeled with dynamin antibodies, as viewed by light and electron microscopy (12). In contrast to these observations, immunoblot analysis comparing the distribution of clathrin and dynamin immunoreactivity in subcellular fractions from rat brain does not demonstrate a convincing colocalization (20). Instead, there is a prevalence of dynamin in a heterogeneous membrane pellet and a reduction in the plasma membrane-coated vesicle fraction. Furthermore, of the numerous antibodies made to the conserved and isoform-specific domains of the dynamins, only one to date (12) has been demonstrated to localize with clathrin in mammalian cells. Thus, while the majority of data supports the involvement of the ubiquitously expressed DynII in CME, it is also possible that dynamins are involved with other non-clathrin-mediated cellular events.

The role of DynII in other endocytic processes, such as fluid phase endocytosis (FPE) or macropinocytosis, is less clear due partly to our incomplete understanding of these cellular events. For example, numerous studies have indicated that both FPE and macropinocytosis are distinct pathways that do not require clathrin, while others suggest that these processes are either linked or fully dependent upon CME (21, 22). Insights into the contribution of dynamin to FPE have been provided by studies on shi^ts flies, either in the intact organism (3, 5, 23) or in cells isolated in culture (24). In both models, the restrictive temperature prevents internalization of fluid phase markers, such as horseradish peroxidase (3, 5), fluorescent plasma membrane lipids, and dextran (24). These observations are interesting because they contrast with the studies described above in transfected cells, which were defective in CME but not FPE (10-13). This suggests that these two endocytic processes are not intimately linked, as some have predicted, and highlights a significant contrast between the different experimental systems. Such a discrepancy could be accounted for by differences in the origin of the cells examined (fly neurons versus mammalian epithelium) or an incomplete inhibition of endogenous dynamins by the overexpressed mutant forms. It is possible that a DynII-independent mechanism of FPE may be increased in cultured mammalian cells to compensate for the inhibition of CME. Indeed, analysis of cultured cells overexpressing a temperature-sensitive mutant of dynamin (13) show a significant decrease in FPE (>50%) at early time points after shifting to the restrictive temperature. At later time points, FPE recovers and is increased to levels beyond that observed in control cells. This suggests that CME is responsible for a significant portion of the total FPE in a cell. However, once CME is inhibited, an additional clathrin- and dynamin-independent pathway is amplified to compensate.

While DynII is likely to have an important endocytic role in all cells, it is reasonable to predict that DynI and DynIII, which are generally limited to either brain or testis and brain, respectively, support more specialized membrane trafficking events in these tissues. In the testis, DynIII has been localized to Sertoli cells, which surround and "nurse" the developing spermatogonia through an active and elaborate endocytic processes (17). These Sertoli cells also absorb residual cytoplasm discarded by the nearly mature sperm. Thus, it is not surprising that this tissue may require the action of a second isoform. Similarly, neurons possess unique endocytic mechanisms that may use DynI to support the rapid and efficient retrieval of synaptic plasma membrane following synaptic transmission.

Because the mechanisms by which SVs recycle are undefined, the participation of DynI in this process is unclear. There are at least three current models for SV recycling. Two models implicate clathrin-mediated vesicle budding, either with or without an endosomal intermediate, while a third model describes a clathrin-independent mechanism by which SVs establish a "fusion pore" to incompletely fuse with and then release from the presynaptic membrane. This "kiss and run" model (25) is attractive, since it provides a mechanistic explanation for how SVs might recycle rapidly without having to take on the more laborious and time-consuming task of assembling clathrin baskets. In addition, a fourth recycling model has been provided by recent electrophysiological studies of endocytosis in cultured pituitary cells demonstrating that limited surface membrane can be retrieved rapidly (<4 sec) after secretion into structures that are 10 times larger than coated vesicles (26). Thus, it is likely that multiple vesicle recycling mechanisms are used in neurons, and they depend on the cell type and the form or intensity of stimulation.

It is unclear which of the endocytic mechanisms listed above are dependent upon the action of both dynamin and clathrin. An electrophysiological study measuring changes in membrane capacitance during stimulated secretion in isolated adrenal chromaffin cells has shown that rapid endocytosis, but not exocytosis, is ablated when either one of two different purified dynamin antibodies are included in the patch pipette (27). Interestingly, manipulations that normally disrupt clathrin function, such as potassium depletion, changes in pH, and clathrin antibodies, had no effect on rapid endocytosis in these cells. In contrast, a recent study by Takei et al. (20) has provided extraordinary electron microscopy images showing both clathrin and dynamin on membranous invaginations in isolated rat brain synaptosomes treated with GTP[S]. In these preparations, one can clearly see densely stained, regularly spaced striations around the neck of the long invaginations (Fig. 3 b-d). These striations label with a monoclonal antibody to dynamin, while clathrin antibodies stain some of the bulbous vesicular ends, suggesting a proximal, if not intimate, relationship between the two proteins. The formation of these striated tubules is remarkably similar to the collared pits observed in the nerve terminals of shi^ts flies at the restrictive temperature (Fig. 1 a and c). In a recent extension of this in vitro study (28), striated dynamin-coated membrane tubules are also seen on internalized endosomal compartments that are either continuous or discontinuous with the presynaptic plasma membrane. These findings support the first two models for SV recycling described above, which implicate an endosomal intermediate from which SVs recycle via a clathrin-based budding mechanism. There is an inconsistency, however, between these observations in regard to the relationship between clathrin and dynamin. In shi^ts flies, the vast majority (>95%) of the presynaptic plasmalemmal invaginations seen at the restrictive temperature have putative dynamin collars yet are not clathrin-coated. In dramatic contrast, all of the pit structures observed in the fly oocytes and nephrocytes are coated yet lack collars (refs. 5 and 6; J. H. Koenig, personal communication). These are reminiscent of the coated structures that lack collars observed in the transfected cultured mammalian cells expressing mutant dynamins (12). Whether the regulated mechanisms for rapid membrane recycling in neurons require DynI, but not clathrin, while the slower constitutive uptake of cell surface receptors, such as those for low density lipoprotein, polymeric IgA, and transferrin, use both clathrin and DynII needs to be tested. However, the shi^ts mutant phenotype is presumed to be the result of a single aberrant dynamin gene product (2, 8, 9) and thereby makes the distinct neuronal and epithelial phenotypes more difficult to explain in D. melanogaster.

Dynamin Interaction with Multiple Subcellular Components

Numerous biochemical studies have indicated that dynamins bind to several different cellular proteins (29). Pleckstrin homology and proline-rich domains in the primary sequence of the dynamins (19) implicate them in tyrosine kinase receptor signal transduction pathways (30). These two distinct domains, which can putatively mediate protein-protein interactions, suggest that the dynamins may be involved in cell growth and differentiation. In addition, the carboxyl-terminal region has been shown to bind to a number of other subcellular components.

The first dynamin-associated protein to be identified was tubulin. Indeed, brain dynamin was originally isolated by its nucleotide-sensitive interaction with microtubules (Mts; ref. 31), a property usually exhibited only by true Mt-based motor enzymes. While the evidence supporting a biological interaction between dynamin and Mts is both substantial and provocative, it is far from absolute. Electron microscopy studies show that dynamin binds and crosslinks Mts (refs. 14, 31, and 32; Fig. 4 e and f). This crosslinking appears to represent a specific interaction based on a conserved periodicity that is exhibited by the well-characterized conventional Mt-associated proteins. Furthermore, like a true Mt-associated protein, dynamin promotes the assembly of purified sea urchin egg tubulin in vitro (34). Finally, tubulin polymer stimulates a substantial 75-fold increase in the GTPase activity of brain dynamin (refs. 31, 36, and 37; Table 1). This stimulation appears to be specific, since it is not mimicked by other acidic cytoskeletal proteins, such as actin, clathrin, and vimentin (31). Despite these observations, the physiological relevance of the interaction between tubulin and the dynamin proteins remains controversial for several reasons: first, it is not clear if Mts do (27, 38) or do not (39, 40) support the early steps of endocytosis; second, dynamin has not been found to colocalize with Mts in vivo (14, 41); and third, cultured cells transfected to overexpress mutant forms of dynamin do not show any alteration in Mt organization (10, 12). Certainly future in vivo studies on the interaction of dynamin with Mts will prove informative.

Table 1. Effect of SH3 domain-containing fusion proteins on dynamin GTPase activity Addition GTPase activity, mol/min per mg None 9.9 ± 2.0 GST 17.0 ± 0.1 c-Src SH3 85.0 ± 12.4 GRB2 131.9 ± 11.9 GRB2 SH3 (N) 16.4 ± 0.9 GRB2 SH3 (C) 12.0 ± 1.4 Mts 658.2 ± 8.3 Dynamin GTPase activity was assayed in the presence of either the indicated glutathione S-transferase (GST) fusion proteins (0.08 mg/ml) or Mts (0.12 mg/ml). Errors indicate the range of duplicate data points. N, N-terminal SH3 domain; C, C-terminal SH3 domain. [Reproduced with permission from Herskovits et al. (36) (Copyright 1993, Proc. Natl. Acad. Sci. USA).]

In addition to Mts, both acidic phospholipids and endogenous rat brain vesicles stimulate the GTPase activity of brain dynamin through an interaction with the basic carboxyl terminus (37). Detailed analysis of this phospholipid-stimulated activity revealed positive cooperativity with regard to enzyme concentration, indicating that the interaction of more than one dynamin molecule is necessary for the full activation promoted by phospholipids (42, 43). As discussed later, it is attractive to speculate that dynamin binding to the plasma membrane, concomitant with self-assembly, activates this enzyme to sever nascent endocytic invaginations.

In addition to a structural mechanochemical function, it has been predicted that dynamin participates in cell signaling cascades. As stated earlier, the carboxyl terminus of dynamin contains a proline-rich domain, which is important for its interaction with Src homology 3 (SH3) domain-containing proteins. Gout et al. (44) have demonstrated by affinity chromatography that the SH3 domains of multiple signaling proteins preferentially bind dynamin from brain extracts. This work also shows that dynamin binds selectively and with high affinity to the SH3 domains of phospholipase C, growth factor receptor-bound protein 2 (Grb2), and the regulatory subunit of phosphatidylinositol-3-kinase (p85), but not to several other SH3 domain-containing proteins. Certain SH3 domain-containing fusion proteins stimulate the GTPase activity of brain dynamin (Table 1). This is likely to be mediated by consensus SH3 domain binding regions located at the carboxyl terminus, which consist of several proline-rich sequences (44). Using synthetic peptides spanning these regions for binding and competition studies, multiple groups have identified several different regions responsible for binding to all of the substrates with the same specificity displayed by the native protein (44-47).

The interaction of dynamin with these various SH3 domain-containing proteins suggests that it is involved in tyrosine kinase receptor-mediated signal transduction. In support of this hypothesis, dynamin has recently been shown to immunoprecipitate with a number of tyrosine kinase growth factor receptors, including the epidermal growth factor receptor (47) and the insulin receptor/insulin receptor substrate I complex (48) via joint interactions with Grb2. Dynamin is complexed with Grb2 in the absence or presence of growth factors, but becomes associated with the respective tyrosine-phosphorylated receptor only upon growth factor stimulation (46, 47). In addition, Scaife et al. (45) report that the association of dynamin with phospholipase C and the platelet-derived growth factor receptor is further increased with growth factor stimulation. These results indicate that dynamin may interact with distinct signaling cascades in response to different agonists. It should be noted however, that like its putative interactions with Mts, dynamin has yet to be localized in vivo with any of the signaling proteins described above. Furthermore, despite the numerous studies done on the interactions of the dynamins with signaling proteins, there are presently no mechanistic or functional studies demonstrating the participation of a dynamin in any cell signaling cascade. We do not know if such a "signaling" function would represent the regulation of a focused, mechanical process (i.e., initiation of endocytic membrane scission) or participation in a more global, classical signaling transduction pathway (i.e., Ca²⁺ regulation or transcriptional activation).

Regulation and Assembly of Dynamin

Despite the numerous studies implicating the dynamin family in endocytosis and signaling, many questions remain unresolved. Mainly, what precise mechanochemical processes do the dynamins support, and how are these functions initiated and regulated? The recent observation that purified DynI, like tubulin and actin, can self-assemble into helical filaments, although in a GTP-independent manner, has provided an important clue into how it might interact with membranes during an endocytic event (ref. 33; Fig. 4 a and b). When combined with the study by Takei et al. (20), which was discussed earlier, it is likely that the electron-dense striations situated on endocytic membranes in the shi^ts flies represent assembled dynamin. Thus, it is attractive to speculate that assembly of dynamin rings around the "neck" of endocytic invaginations is followed by a GTP-dependent contraction which induces membrane severing (33, 49). Consistent with this model, multiple observations demonstrate that an inhibition of dynamin GTPase activity reduces endocytic membrane budding. For example, mutations in the GTP-binding domains of dynamin lead to an accumulation of membrane invaginations in both shi^ts (3-6) and transfected cells (10-13), as does addition of the nonhydrolyzable analogue GTP[S] (20). How and when this GTPase activity is stimulated during endocytosis is not clear, but as discussed previously, it can be increased in vitro by multiple components including dynamin itself (42, 43), Mts (14, 31, 34), membranes (37), and SH3 domain-containing proteins (36, 44). Because these cytoplasmic constituents are situated at or near the invaginating endosome, they are likely to be true biological activators.

It is also likely that the functions of the dynamin proteins can be modulated further through posttranslational modifications such as phosphorylation. In vitro studies have demonstrated that protein kinase C phosphorylates DynI and leads to a stimulation of GTPase activity, while the calcium-dependent phosphatase calcineurin can reverse this stimulation (50-52). These studies show that only DynI appears to be a substrate for these enzymes and provides the first evidence supporting differences in function between the dynamin isoforms. These biochemical in vitro observations are further supported by a study using isolated synaptosome preparations, which suggests that phosphorylation may alter interactions between the dynamins and membranes. These observations show: (i) brain dynamin predominantly associates with membranes in a dephosphorylated form, while a small percentage remains cytosolic and phosphorylated (51); (ii) activation of calcineurin during calcium-induced depolarization of nerve terminals corresponds with the dephosphorylation of DynI (52); and (iii) the action of calcineurin reverses the effects of protein kinase C phosphorylation and inhibits the DynI GTPase activity, allowing it to reassociate with synaptic membranes (52). Taken together, these data support a model in which DynI cycles from cytosol to membrane during synaptic transmission, yet it is unclear how to interpret these findings in relationship to those described above. For example, the observation that phospholipids and membranes stimulate the GTPase activity of brain dynamin (42) appears to conflict with observations that an activated phosphorylated dynamin remains cytosolic. Certainly, future studies which define the synergism between dynamin phosphorylation and its interaction with other cellular components will provide insight into how this enzyme participates in membrane budding events.

Multiple Functions for the Dynamin Family

It is well documented that the dynamins are members of an extended family of large GTP-binding proteins (53), which include the yeast vacuolar protein sorting protein, Vps1p (54), the Mx subfamily, which confers cells with selective resistance to viral infection (55), and the mitochondrial protein MGM1 (56). While our understanding of the cellular functions performed by MGM1 and Mx are limited, recent studies on Vps1p mutants have proven informative. Vps1p possesses an amino terminus that shares 66% homology to the dynamins and contains the tripartite GTP-binding motif, while the carboxyl terminus is believed to convey targeting to specific membrane compartments (57). These compartments are likely to include the trans-Golgi and the vacuole, since Vps1p mutants missort both resident Golgi endoproteases and vacuolar membrane proteins to the plasma membrane where they are reendocytosed and indirectly sent to the vacuole (58, 59). It is unclear whether a homologue is present in mammalian cells or if Vps1p is the yeast dynamin equivalent. Although searches for a mammalian Vps1p gene have proven inconclusive to date, another dynamin-like gene (DNM1) has been recently identified in yeast (60). Although this gene product appears to be involved in an endocytic process, it is more similar to Vps1p, since these proteins do not contain a pleckstrin homology domain or the proline-rich carboxyl terminus characteristic of the mammalian dynamins.

While substantial experimentation indicates that dynamins function at the plasma membrane, it is enticing to predict that a dynamin isoform or family member could participate in membrane dynamics during other membrane trafficking processes (Fig. 5). It has become increasingly clear that endocytosis and secretion share common mechanisms that support membrane budding and fusion. Indeed, multiple proteins, such as the adaptins, clathrin, caveolin, and the ARF family, have been implicated in processes at both the Golgi and plasma membrane (62-67). Therefore, it is not surprising to learn that multiple antibodies made to the conserved amino-terminal domain of the dynamins stain the Golgi apparatus in cultured cells, as do antibodies to yeast Vps1p (61). It is certainly attractive to speculate that either a dynamin isoform or related protein acts as a "pinchase" at the trans-Golgi surface to liberate nascent, clathrin-coated secretory vesicles. This prediction is inconsistent, however, with observations made in isolated cells of shi^ts flies or cultured mammalian cells transfected with mutant DynI. In these systems, endocytosis is perturbed, yet several specific secretory processes are unaltered, including normal trafficking of VSV-G protein or fluorescent Golgi lipids to the plasma membrane (24), transport of nascent lysosomal proteins (11), and localization of the Golgi-associated -adaptin (10). It is possible that some processes in these manipulated cells are more sensitive to altered dynamins than others, as demonstrated by the differential effects on CME versus FPE (10, 12, 13). Furthermore, these transfection studies may not affect all of the many dynamin-like proteins in a cell. Certainly, future functional studies testing the role of dynamins in the secretory pathway will be useful.

In addition to an endocytic function and a possible role in secretion, observations in several different cell models have indirectly implicated dynamin in the regulation and maintenance of cell shape. For example, neurons dissociated from shi^ts larvae are unable to grow neurites or growth cones when maintained at the restrictive temperature (68). In support of these findings, we have demonstrated that a reduction in the intracellular level of DynI in cultured rat hippocampal neurons through antisense oligonucleotide treatment results in a significant impairment of neurite formation (69). In addition, we have found that DynI gene expression is increased during neurite growth in culture, while that of DynII remains unchanged (15). Finally, cultured mammalian cells that have been transfected to overexpress mutant DynI show marked alterations in shape and cytoskeletal organization (12). Whether these phenomena are induced by direct perturbations in cell signaling cascades or cytoskeletal dynamics or, alternatively, simply represent the indirect effects of an impaired endocytic pathway remains to be determined.

Conclusion

Here we have provided an overview of an expansive number of studies that lend insight into the diversity, localization, function, and enzymology of the dynamin family of large GTPases. It is now clear that dynamins are expressed in all cells and perform an important and universal function during endocytosis. There are three closely related dynamin genes and corresponding products that are alternatively spliced. These comprise one component of a broad superfamily of tissue-specific proteins that support membrane budding and trafficking events. Despite the many advances in this field, important questions remain to be addressed. There are likely to be other dynamin genes yet to be identified, and the distributions and functions of these isoforms in various cell types need to be defined. Finally, whether the function of the dynamins extends beyond endocytosis to participate in cell signaling cascades or other membrane budding events remains to be established.