Kinesin-II Is Required for Axonal Transport of Choline Acetyltransferase in Drosophila

Unless noted, all molecular biology experiments were done by standard methods (Sambrook et al. 1989). Fly husbandry and other related procedures were as described (Ashburner 1989; Goldstein and Fyrberg 1994). All fly stocks, unless otherwise mentioned, are described in Lindsley and Zimm 1992.

A 150-bp PCR amplified KLP64D (KLP4 in Stewart et al. 1991) sequence was used as a probe to isolate several truncated cDNA clones from a Drosophila 0–4-h embryonic library cloned in pNB40 (Brown and Kafatos 1988). A composite 2.5-kb KLP64D coding sequence was created from several cDNA clones. Nested deletions were created in the cloned cDNA fragments in either direction using an ExoIII kit (Promega Corp.) according to the protocol provided by the manufacturers. Both strands were sequenced using either the T7 or the T3 primers, and later the whole parental clones were sequenced using the internal primers generated from the primary sequences. Several overlapping genomic fragments were isolated from a Drosophila genomic library in Lambda DASH^® II using the KLP64D cDNA probe, and the coding sequences were further mapped to a 3.6-kb EcoRI/HindIII fragment, which was then cloned into pBluescript SK+ (Stratagene Inc.) and called pB11-4. Sequencing this genomic fragment revealed a single continuous open reading frame without any intron, and matched the KLP64D cDNA sequence.

Several deletions in the KLP64D gene were isolated previously by imprecise excision of P-transposable elements inserted at the 3′ end of the gene (Perez and Steller 1996). All mutant chromosomes were maintained over a TM3, Ser <y+> balancer in a yw background. We generated point mutations in the KLP64D gene by scoring failure of ethyl methanesulphonate (EMS)-induced lethal mutations to complement a deletion that removes the KLP64D gene, Df(3L)Klp64D^A8.n123. In brief, mutagenesis followed standard methods (Grigliatti 1986) in which 2-d-old adult isogenic w males were fed with 27 mM EMS in 2% sucrose solution for 16 h, put on fresh food for 1 d, and then groups of 10 males were mated to 40 w; TM3, Sb Ser/TM6B, Tb Hu (Lindsley and Zimm 1992) virgin females (60 matings). Individual male progeny that were TM3, Sb Ser/* or TM6, Tb Hu/* (where * indicates a mutagenized third chromosome) were then mated to four yw Df(3L)Klp64D^A8.n123/TM3, Ser <y+> virgins (9,000 matings). The progeny of these matings were screened for the absence of */Df(3L)Klp64D^A8.n123flies. TM3,Ser /* males from appropriate lines were then mated to yw, TM3, Ser <y+>/MKRS Sb virgins to create a stock. The mutagenized third chromosome was cleaned up and the lethal mutations mapped by recombination with a chromosome containing ru h th st cu sr e ca. We found eight recessive lethal mutants that failed to complement Df(3L)Klp64D^A8.n123 and that mapped to the KLP64D region. These lethal mutations fell into three complementation groups, one of which was found to identify the KLP64D gene.

A 3.6-kb KLP64D genomic fragment in pB11-4 was cleaved out by NotI and XhoI endonuclease digest and ligated to the same sites of pUAST (Brand and Perrimon 1993) and named pUAS64D. This maintained proper orientation of the KLP64D coding sequences with respect to the UAS sites in pUAST. We recovered several stable transformant lines (marked by w+ eye) in the X and in the second chromosome by injecting pUAS64D with a helper plasmid, containing the P-transposase gene, in yw embryos. These lines were then crossed to the Klp64D mutants. Rescue of lethality was scored both at the larval stages as well as in the adult, where the homozygous mutants are marked by yellow body color.

For rescue experiments using the mouse KIF3A gene, an inducible transgene was created by inserting a full length KIF3A cDNA (Kondo et al. 1994) obtained by standard methods between the EcoRI and the NotI restriction sites of the pUAST vector (Brand and Perimon, 1993). This arrangement placed the 5′ end of the KIF3A coding sequence near the GAL-UAS transcription activation sites. Stable transformant lines were then obtained by P-element–mediated transformation. Expression of the KIF3A transgene was induced in all neurons by Gal4 driven from the promoter of the elav gene inserted in the third chromosome (Luo et al. 1994).

In situ hybridization was done according to Tautz and Pfeifle 1989, with modification as described in Pesavento et al. 1994 and Perez and Steller 1996. An antisense strand-specific ribo-probe was made from the 1.6-kb Spe I/Hind III fragment subcloned in pBluescript SK.

An affinity column was prepared by cross-linking about 10 mg of truncated, bacterially expressed and purified glutathione S-transferase–KLP68D (Pesavento et al. 1994) to 1 ml Affigel-10 (Bio-Rad Laboratories) beads according to the protocol supplied by the manufacturer. These were packed in a 15-ml plastic column (Bio-Rad Laboratories) and blocked thoroughly with 100 ml of 10% BSA (Sigma Chemical Co.) in 10 mM Tris-Cl, pH 7.0, 100 mM NaCl, and 5 mM KCl (TBS-BSA), then washed in succession with 15 ml of 100 mM glycine, pH 2.5, TBS-BSA, 100 mM triethanolamine, pH 11.5, and finally twice with TBS-BSA. It was then loaded with the antiserum, washed twice with 15 ml of TBS-BSA, and the bound antibody was extracted with 1 ml of 100 mM glycine, pH 2.5 (Harlow and Lane 1988). We tried all three methods of 1 M MgCl₂, 100 mM glycine, and 100 mM triethylamine extraction and found glycine to be the most suitable. The purified antibody was dialyzed against TBS overnight and stored at 4°C with 0.2% NaN₂.

For Western blotting, ∼50 heads, thoraces, and abdomens from male and female adult flies were dissected and collected separately in plastic microcentrifuge tubes on ice. They were squashed in 500 μl sample loading buffer containing 0.1 M β-mercaptoethanol, 10% glycerol, and 0.1% bromophenol blue, and centrifuged at 14,000 g for 5 min at room temperature. The supernatant was boiled for 3 min and then separated by 10% SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Harlow and Lane 1988). The Western blots thus prepared were stained with several primary antisera and corresponding peroxidase-coupled secondary antisera (Jackson Laboratories, Inc.), and developed using the ECL kit (Amersham Pharmacia Biotech) according to the protocol supplied by the manufacturer. For coimmunoprecipitation experiments, methods were essentially as described in Yang and Goldstein 1998. In brief, extracts of isolated Drosophila heads in RIPA buffer (50 mM Tris-Cl, pH 8.0, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, 150 mM NaCl) were subjected to immunoprecipitation, run on SDS polyacrylamide gels, and then probed in Western blots by standard methods.

Embryos were dechlorinated in 50% commercial bleach for 3 min and washed thoroughly in running distilled water, then fixed in heptane–fixative (2:1) interface as described in protocol 90 (Ashburner 1989) for 20 min at room temperature, and then devitelinized using methanol. The fixative contained 4% formaldehyde (EM grade) in 100 mM Pipes, pH 6.9, 1 mM EGTA, and 2 mM MgSO₄. Larval tissues were dissected in Drosophila Ringers containing 130 mM NaCl, 4.7 mM KCl, 1.8 mM MgCl₂, 0.7 mM KH₂PO₄, and 0.35 mM Na₂HPO₄ and then fixed in the same fixative as mentioned above for 20 min at room temperature.

Fixed tissue samples and embryos were washed with PBS and 0.1% Triton X-100 (PTX) and incubated with the primary antibody solution in PTX for 1 h. They were then washed with five changes of PTX and incubated with fluorescent secondary antibody (1:250 dilution in PTX; Jackson Laboratories, Inc.) for 1 h. Finally, after another wash with five changes of PTX, they were mounted with a drop of Vectashield™ mounting media (Vector Laboratories, Inc.). We used goat anti-HRP (1:1,000 in PTX; Cappel Inc.), purified rabbit anti-KLP68D (1:5 in PTX; Pesavento et al. 1994), mAb anti-ChAT (1:1,500 in PTX; Kitamoto et al. 1995), rabbit anti-synaptotagmin (SYT) (1:200 in PTX; Littleton et al. 1993), and mAb anti–cysteine string protein (CSP) (1:50 in PTX; Zinsmaier et al. 1994). The secondary antibodies alone do not stain anything in the control tissue.

The stained tissue samples were observed using a Bio-Rad MRC1024 laser scanning confocal device which was fitted to a Nikon Diaphot 200, and the images were captured by the Laser-Sharp 3.0 software. Images were further processed in Adobe Photoshop 4.0 for presentation. Most of the pictures contain data from 3–10-μm thick regions in the tissue.

KLP64D (previously named KLP4) was identified during a PCR screen for new Drosophila genes with homology to the highly conserved regions of the KHC motor domain (Stewart et al. 1991). The gene was mapped to salivary gland chromosome band 64D. Northern analysis revealed a 2.6-kb mRNA enriched in the head (Stewart et al. 1991). Sequence of the full-length KLP64D coding region was obtained from several cDNA clones as well as a genomic clone. Analysis of the sequence indicated a 2032-bp open reading frame (ORF), which encodes a 674–amino acids (aa) long polypeptide with a predicted molecular mass of 76.4 kD. Secondary structure prediction analysis suggested three distinct domains in the polypeptide, including a globular NH₂-terminal domain of 385 aa, an alpha-helical coiled-coil central domain of 200 aa, and a globular COOH-terminal tail domain of ∼100 aa. The KLP64D sequence is most similar to the kinesin-II subunit KIF3A (94% similarity) of mouse and KRP85 (93% similarity) of sea urchin (Fig. 1). Considerable similarity (66%) was found with another Drosophila kinesin-II polypeptide encoded by the KLP68D gene (Pesavento et al. 1994). Therefore, we placed KLP64D in the kinesin-II family with KIF3A and KRP85 as the likely homologues. This assignment is also supported by the finding that the lethality of KLP64D mutants can be rescued by expression of mouse KIF3A (see below).

To obtain clues about the potential functions of kinesin-II in neurons, we examined the expression patterns of KLP64D and KLP68D in the developing Drosophila nervous system. In overview, we found that KLP64D and KLP68D are both expressed in a restricted set of neurons, which appear to correlate with the cholinergic subset.

In situ hybridization to RNA revealed that KLP64D is expressed in neurons of both the central (CNS) and peripheral (PNS) nervous system. KLP64D RNA was first observed in the CNS at stage 12 (Fig. 2 A, arrows) when many CNS neurons begin extending their processes (Broadie and Bate 1993; Goodman and Doe 1993; Campos-Ortega and Hartenstein 1997). Later in embryogenesis, at stage 16, KLP64D expression continues in the CNS, including the optic lobes (Fig. 2 B, large arrow). Prominent expression was also found in the PNS, where KLP64D RNA was observed in the cells of the dorsal (Fig. 2 B, arrowheads) and lateral chordotonal organs (Fig. 2 B, small arrows). Careful observation revealed that the staining in the abdominal lateral chordotonal (lch5) organs (Fig. 2 C, arrows) was restricted to the neuronal cells. In addition, we also found staining in the ventral and dorsal chordotonal neurons, the anterior sense organ and in the antenno-maxillary complex (data not shown). Expression in the CNS and PNS continues in the first instar larva and was still detectable in the third instar larval brain, although at a lower level (Fig. 2 D). Expression in the third instar brain includes cells of the ventral ganglion (Fig. 2 D, arrowheads) and a few cells in the optic lobe (Fig. 2 D, arrow).

Antibody staining revealed preferential expression of KLP68D in a subset of neurons similar to those shown to express KLP64D, and shown previously to express KLP68D RNA (Pesavento et al. 1994). The antibody used was an affinity-purified antiserum raised against the 432-aa COOH-terminal domain of KLP68D (Pesavento et al. 1994). In Western blots of adult head lysate, the purified serum stained a single band of 90 kD, which is close to the expected molecular mass of full-length KLP68D. In stage 17 embryos, the purified KLP68D antibody strongly stained a subset of PNS neurons, including chordotonal neurons, cells in the antenno-maxillary complex (Fig. 3A, Fig. C, and Fig. E), bolwig organ cells (Fig. 3 A, asterisk), and other sensory neuron cells in the epidermis. To visualize all neurons, the same embryos were also stained with anti-HRP antibody (Fig. 3B, Fig. D, and Fig. F), which, as expected, labeled the cytoplasm and membranes of all neuronal cell bodies and processes (Jan and Jan 1982). This double staining experiment helped establish the identity of the KLP68D-positive sensory neurons in the epidermis (Fig. 3), and revealed that KLP68D may be specific to a subset of PNS sensory neurons, whose common thread may be that they are all cholinergic (see below).

KLP68D was also detectable in the CNS starting at stage 16 in the embryo, where it was most pronounced in the first instar larva stage in the ventral ganglion as well as in the optic lobe. Unlike the sensory neurons, the expression in the CNS was not distinguishable in specific cell types, owing to the small size and high density of cell bodies in this region. Some cells in the optic lobe were strongly stained early in the first instar stage, but the staining was not seen at a later stage. In the ventral ganglion, uniform staining appeared in the neuropil and cell cortex. No specific staining was detectable in CNS at the third instar stage.

A comparison of the KLP68D and KLP64D expression patterns (Fig. 2) suggests that both KLP64D and KLP68D are expressed in similar tissue subsets and most likely in the same set of cells. However, the overlap could be determined with certainty only in the neurons of the chordotonal organs (Fig. 2 C and Fig. 3 E), bolwig organ cells, and in the antenno-maxillary complex. We noted that the KLP64D and KLP68D distribution was similar to that of acetyl cholinesterase RNA expression (Zador 1995) and ChAT RNA and lacZ reporter expression in the PNS (Yasuyama et al. 1995), which are good markers for cholinergic presynaptic neurons in Drosophila and other insects. The neurons innervating the chemosensory, chordotonal, and larval bolwig organs (Kitamoto et al. 1995; Yasuyama et al. 1995) are known to be cholinergic. In addition, many interneurons of the CNS also express ChAT and acetyl cholinesterase (Yasuyama et al. 1995; Zador 1995). Unlike vertebrates, motor neurons in Drosophila generally use glutamate instead of acetylcholine as their neurotransmitter (Johansen et al. 1989). Similar to what others have found, our efforts to localize ChAT antigen itself in the neuronal cell bodies of the PNS and CNS using ChAT antiserum were not successful, and so we could not determine the overlap of ChAT and KLP68D in the cell bodies with precision. However, since all of the cholinergic synapses are located in the CNS, ChAT staining was detected in the synaptic bulbs of the CNS neuropil (see Fig. 5 A), which provided an assay for ChAT transport in KLP64D mutants (see below).

To test further whether KLP68D and ChAT are coexpressed in cholinergic neurons, we used kinesin light chain (Klc⁻) mutant larvae. Homozygous Klc mutant larvae die at the late second instar and early third instar stages with progressive distal paralysis. The axons of these larvae develop focal swellings along their length and contain accumulated cargos of various types that appear to block single axons in axon bundles (Hurd and Saxton 1996; Gindhart et al. 1998). We observed numerous KLP68D and ChAT immunoreactive clogs in the axons of these animals (Fig. 4A and Fig. B, arrows), which probably result from the blockage of ChAT and KLP68D transport in cholinergic neurons. Quantitation in several nerve bundles revealed strong coincidence between ChAT immunoreactivity and that of KLP68D in the axons (Table). This observation confirms the suggestion that both ChAT and KLP68D are expressed in the same cells and transported in the same axons. These findings are significantly different than observed for ChAT and a ubiquitous axonal cargo, SYT (Fig. 4, C–F; Table). Although there are a large number of SYT-positive clogs in the axon, only a small percentage (37.4%) are also ChAT-positive (Table; Fig. 4E and Fig. F, arrows). These data suggest that not all axons and axonal clogs contain ChAT and therefore, that the observed ChAT and KLP68D association is not fortuitous.

Since other members of the kinesin-II family are often associated in heterodimeric complexes, we conducted coimmunoprecipitation experiments to probe the association of KLP68D and KLP64D. These experiments were complicated by the absence of an antibody specific for KLP64D and the lack of cross-reactivity in Western blot experiments of antibodies reactive with the KIF3A and KRP85 homologous proteins (data not shown). However, it was possible to express functional mouse KIF3A in Drosophila using the GAL4-UAS system (Brand and Perrimon 1993), and to demonstrate that KIF3A rescues KLP64D mutants (see below). Flies expressing KIF3A could then be used in coimmunoprecipitation experiments. Immunoprecipitations using the K2.4 mAb raised against sea urchin KRP85 (Cole et al. 1993) precipitated both KIF3A and KLP68D in flies expressing KIF3A, but not in wild-type (Fig. 5). Complementary immunoprecipitations using anti-KLP68D also precipitated both KIF3A and KLP68D in flies expressing KIF3A, but not in wild-type (Fig. 5). Control immunoprecipitations precipitated no immunoreactive material in any genotype. Thus, we propose that KLP64D, which is the sequence and functional homologue of KIF3A, also associates with KLP68D in vivo.

Previous work led to the generation and characterization of a set of deletions in and around the KLP64D gene (Perez and Steller 1996). These deletions were obtained by imprecise excision of several P-transposable elements inserted near the 3′ end of the KLP64D gene (Perez and Steller 1996). One such deletion, Df(3L)Klp64D^A8.n123, has a left breakpoint that extends to the extreme 5′ end of the KLP64D protein coding sequence. To generate KLP64D mutations, we used EMS mutagenesis and screened for lethal mutants that failed to complement Df(3L)Klp64D^A8.n123. We isolated eight mutants that fell into three different complementation groups. Meiotic recombination mapping with respect to roughoid and hairy placed the lethality in all of these chromosomes near the KLP64D gene. One of the complementation groups containing three of the EMS alleles, Klp64D^k1, Klp64D^k5, and Klp64D^k33, identifies the KLP64D gene, since these mutants show changes in the KLP64D coding sequence. Specifically, Klp64D^k1has a change from Gln (CAA) to a stop (UAA) at codon 13, Klp64D^k5has a change from a highly conserved Glu (GAG) to Lys (AAG) at codon 551, and Klp64D^k33 has a change from T to G in the 3′ untranslated region 4 bp after the predicted normal stop codon. We presume that Klp64D^k33 causes mRNA instability, since its phenotype is similar to that of Klp64D^k1, which is likely to be a null-mutant since it has a stop codon early in the gene. Heteroallelic combinations among these mutants give some complementation and a few very uncoordinated and near-paralyzed adults.

All three KLP64D mutants are homozygous lethal. In a typical crowded culture, these mutants die at or before the early third instar larval stage. If the mutants are raised in conditions of low density and without competition from wild-type larvae, a few mutant animals will survive to be pharate pupae and occasionally eclose as fully formed adults. All of these mutant adults exhibited an acutely uncoordinated walk, are unable to stand, and died in a few hours. Mutant third instar larvae exhibit slight to severe sluggishness, and roll abnormally from side to side during crawling. These observations suggest a neuronal function of the KLP64D gene.

To confirm that mutations in the KLP64D gene were responsible for the lethality and other behavioral defects, we used genetic rescue experiments. Although large KLP64D genomic constructs were unstable in bacteria, we were able to construct pUAS64D, which contained an ∼3.6 kb EcoR1/HindIII genomic fragment with a full-length KLP64D coding sequence, 447 bp of 5′ upstream sequences, including the TATA box, and about 0.8 kb of 3′ sequences in pUAST (Brand and Perrimon 1993). We reasoned that expression of KLP64D from this construct would be initiated from its own promoter if the necessary enhancer element is present within the limited 5′ upstream region. Alternatively, the GAL4-UAS at the 5′ end can be used to drive KLP64D expression through the tissue-specific expression of specific promoter–Gal4 fusion genes (Brand and Perrimon 1993). We obtained several independent stable transformants and crossed them to various Klp64D alleles to test whether they can rescue the lethality and other abnormalities. The pUAS64D construct by itself rescued the lethality and uncoordinated behavior of Klp64D^k1 and Klp64D^k5, either in homozygous condition or hemizygous with the Df(3L)Klp64D^A8.n123 chromosome. These results confirm that the lesions in the KLP64D gene caused the lethality and the uncoordinated behavior.

We also used genetic rescue experiments to establish further that KLP64D is a true functional homologue of KIF3A and KRP85. Thus, a complete KIF3A cDNA was placed under the control of GAL4-UAS sequences and introduced into the Drosophila genome by P-element transformation. Then we combined this transgene with a transgene encoding GAL4 driven by the neural elav promoter (Luo et al. 1994). This combination of elements gave complete rescue of Klp64D^k1 heterozygous with a Df(3L)Klp64D^A8.n123 chromosome. Thus, when yw/Y; UAS-Kif3A<w+>/+; Df(3L)Klp64D^A8.n123elav-Gal4 <w+>/TM3, Ser e <y+> was crossed to yw/yw; +/+; Klp64D^k1/TM3 Ser e <y+>, 92 progeny carrying UAS-Kif3A<w+>/+; Df(3L)Klp64D^A8.n123elav-Gal4 <w+>/Klp64D^k1were recovered, whereas no progeny carrying +/+; Df(3L)Klp64D^A8.n123elav-Gal4 <w+>/Klp64D^k1were recovered out of a total of 414 progeny, which is well within expected proportions.

The expression pattern and the behavioral defect in Klp64D mutant animals led to the hypothesis that the KLP64D motor complex might be involved in transporting the components of the acetylcholine-producing machinery. We tested this view by staining Klp64D mutant animals with ChAT antiserum. In wild-type third instar larvae, we found ChAT in abundance at many synapses in the neuropil of the ventral nerve chord and in the optic lobes of the larval brain (Fig. 6 A), whereas very little was observed in the cell bodies of the brain cortex. Counter staining with rhodamine-labeled α-bungarotoxin, which binds to the nicotinic acetylcholine receptors (nAChR) (Schloss et al. 1992), showed that the ChAT antigen is distributed in the neuropil similar to nAChR (Fig. 6 E). No ChAT immunoreactivity was observed in the axons (Fig. 6 C) of wild-type larvae. This staining pattern is consistent with previous reports (Yasuyama et al. 1995). In Klp64D^k1 and Klp64D^k5homozygous third instar larvae, we observed accumulation of ChAT immunoreactivity in the cell cortex (Fig. 6 B, arrowheads) of the ventral ganglion. Closer examination of these mutant samples revealed considerable ChAT immunogen in the nerve roots (Fig. 6B and Fig. D, arrows) and in the cell bodies (Fig. 6 B, arrowheads). In some samples, the ChAT immunoreactivity in the neuropil was reduced strikingly (data not shown) relative to the wild-type control. The abnormal staining pattern suggests that loss of KLP64D affects ChAT transport to the synapse, since no such ChAT staining is observed in the wild-type axons (Fig. 6 A, arrows). The α-bungarotoxin (Schloss et al. 1992) staining of nAChR on the postsynaptic membrane in the Klp64D^k1 homozygous mutants (Fig. 5 F) showed no detectable difference from wild-type (Fig. 6 E) control.

We tested whether the loss of ChAT immunoreactivity at the synapse and the abnormal accumulation in the axons are a secondary effect of the lethality or other neural defects in the mutant animals lacking KLP64D. We examined Khc⁻ and Klc⁻ larvae, which die at the second instar and third instar larva stage, respectively, with significant axonal transport defects (Hurd and Saxton 1996; Gindhart et al. 1998). We found that the intensity of ChAT staining in late second instar Khc⁻ as well as in late third instar Klc⁻ brains was not significantly reduced or altered as compared with the wild-type. Hence, it is unlikely that reduction of ChAT immunoreactivity is a secondary effect of lethality or general axonal transport defects.

To test whether ChAT accumulation in Klp64D^k1axons is a specific effect of KLP64D malfunction or a result of general axonal clogging, we stained Klp64D^k1 homozygous larvae at the third instar stage with antiactin (Fig. 6C′ and D′), anti-SYT (data not shown) (Littleton et al. 1993), and anti-CSP (Fig. 6G and Fig. H, and Fig. 5 H; Zinsmaier et al. 1994). We used Klp64D^k1/TM3 larvae as the wild-type control (Fig. 6 G). Since both SYT and CSP are integral parts of synaptic vesicles, they are transported to all chemical synapses and are likely to accumulate in general axonal clogs as seen in Khc and Klc mutants (Gindhart et al. 1998; Hurd and Saxton 1996). We found no accumulation of SYT and CSP in the axons of the Klp64D mutant larvae (Fig. 6 H). In addition, we also stained the larvae with anti-HRP, antitubulin, and antiactin (Fig. 6 D′) and found no detectable accumulation in the axon. Finally, we also noted that ChAT accumulation in Klp64D mutant axons is distinct from the clogs found in Klc or Khc mutants (see Fig. 4), since these accumulations do not cause obvious swelling of the axons. In fact, EM observation suggests that general CNS neuronal and axonal morphology is normal in Klp64D mutants (data not shown). Therefore, mutations in the Klp64D gene appear to cause selective accumulation of ChAT as opposed to general clogging. In fact, these results suggest that there are probably no clogs in these axons, which is dissimilar to Klc and Khc mutants.

In this paper we report two principal findings. First, we found that KLP64D and KLP68D, which appear to be mutually associated motor polypeptides of the kinesin-II family, are expressed preferentially in cholinergic neurons in Drosophila. Second, we found that mutations in the KLP64D gene cause accumulations of ChAT in neuronal cell bodies and axons. Since other apparently generic axonal cargos do not appear to accumulate in these mutants, and since the overall neural morphology of mutant animals is relatively normal, the data suggest that these kinesin motor proteins are required for transport of ChAT, and perhaps a small number of other cargos in axons of cholinergic neurons. There are several important implications of these findings.

A striking conclusion that can be drawn from the preferential expression data and the relative selectivity of the axonal transport defect in KLP64D mutants is that KLP64D and KLP68D in Drosophila are required for axonal transport of a subset of axonal cargos, one of which is ChAT. This conclusion suggests that a common assumption about the general organization of the axonal transport machinery may be incorrect. That is, axonal transport is usually thought of as a neuronal process that uses a general machinery of neuronal motors, adapters, and carriers for transport of both specialized and ubiquitous axonal components (Hirokawa 1998). Consistent with this view, there is good evidence for the existence of a general axonal transport machinery that uses ubiquitous neuronal kinesins such as unc104/KIF1A, KIF1B, and various forms of true kinesin to transport cargos used in all neurons such as mitochondria, synaptic vesicle components, and channels (Hall and Hedgecock 1991; Otsuka et al. 1991; Nangaku et al. 1994; Hurd and Saxton 1996; Yonekawa et al. 1998). However, our data suggest that there also may be specialized transport pathways activated or enhanced in specific classes of neurons. These neuronal classes may be defined by neurotransmitter or other phenotype, and may activate specific motors, adapters, and/or carriers for axonal transport. While kinesin-II in Drosophila (KLP64D and KLP68D) may provide the first recognizable example of such a cell type–specific axonal transport system, there may be others awaiting recognition or discovery.

Although the reported expression patterns of kinesin motor proteins in vertebrate systems do not provide a clear test of the hypothesis that some motors are harnessed in cell type–specific axonal transport pathways, there are a few relevant anecdotal observations in the literature. In particular, KIF3B, KIF3C, and neuronal KHC have been reported to have expression patterns that are not uniform among neurons (Niclas et al. 1994; Muresan et al. 1998; Yang and Goldstein 1998). KIF3B was reported to be expressed in a subset of Purkinje cells (Muresan et al. 1998), whereas neuronal KHC exhibits greatly elevated expression in a subset of neurons (Niclas et al. 1994). However, at present, in the absence of double staining experiments with good markers of neuronal phenotype, it is difficult to gauge whether some kinesins have strict cell type specificity in vertebrate neurons. Nonetheless, it is clear that even ubiquitous neuronal motor proteins, such as neuronal KHC, may not be expressed equivalently in all neuronal types. Although some of this nonuniformity in expression might be a consequence of overall transport requirements in large neurons, some of the heterogeneity might result from the need for cell type–specific axonal transport pathways in phenotypically differentiated neurons.

There is considerable information regarding the function of kinesin-II family members in several organisms. In particular, a variety of data suggest that motors of the kinesin-II family are specialized for the transport of cytosolic, nonmembrane enclosed constituents such as dyneins or central pair components of cilia and flagella. In Caenorhabditis, kinesin-II motors are needed for the function of sensory cilia as demonstrated by the phenotype of osm-3 mutants (Shakir et al. 1993; Tabish et al. 1995). In sea urchins, antibody microinjection experiments demonstrate the requirement for these motors in transport of ciliary components, perhaps those of the central pair microtubules (Morris and Scholey 1997). In vertebrates, motors of this family have been found in the immotile dendritic sensory cilium of the rod photoreceptor, where they could move ciliary or sensory components (Beech et al. 1996) and are required to form early embryonic cilia of the node (Nonaka et al. 1998; Marszalek et al. 1999; Takeda et al. 1999). Finally, in Chlamydomonas, there is good evidence to support the view that kinesin-II motors are needed to drive the movement of intraflagellar particles or rafts of axonemal components such as inner arm dyneins (Kozminski et al. 1993, Kozminski et al. 1995; Walther et al. 1994; Piperno et al. 1996; Piperno and Mead 1997). Together, these data have suggested a requirement for kinesin-II in transport in ciliated cells or neurons with degenerate dendritic sensory cilia. What has been unclear is the possible role of this family of motor proteins in axons of sensory and nonsensory neurons.

Our work in Drosophila suggests that a unifying theme for kinesin-II motors may be that they are specialized to move cytosolic proteins in motile cilia and flagella, modified immotile dendritic sensory cilia, and in certain classes of axons. Since ChAT is generally thought to be a cytosolic protein, its dependence upon kinesin-II in Drosophila for transport may reflect this specialization. There is also evidence that kinesin-II motors may be associated with membranous components (Henson et al. 1995, Henson et al. 1997; Yamazaki et al. 1995; Le Bot et al. 1998; Muresan et al. 1998; Tuma et al. 1998; Yang and Goldstein 1998). Thus, kinesin-II motors could, in principle, move both complexes of cytosolic protein components in axons and cilia as well as interact with membranous vesicles. Alternatively, the complexes of cytosolic proteins could interact intermittently with membranes to give the observed associations, similar to the suggestion that has been made for bead movements on the surface of Chlamydomonas flagellar membranes (Kozminski et al. 1995). An intriguing possibility in this regard is that the combinatorial interactions of kinesin-II polypeptides in obligate heterodimers might allow some forms to associate with soluble complexes and some to associate with membranous vesicles (Muresan et al. 1998; Yang et al. 1997). In this context, it remains formally possible that ChAT in Drosophila is not moved as a soluble protein complex, but instead as a “hitch-hiker” on membranous vesicles. Further work will be necessary to fully resolve these issues.

The transport of soluble proteins such as ChAT is normally attributed to the slow axonal transport machinery, which is driven by an unknown mechanism (Nixon 1998). However, past work is not conclusive on the question of whether ChAT is transported in the fast or slow axonal transport compartment (Malatova et al. 1989; Tandon et al. 1996), and at present there is no evidence as to whether Drosophila has a slow axonal transport pathway. An economical but speculative hypothesis is that although kinesin-II motors are generally thought of as fast axonal transport motors based on their speed of movement in vitro and on the velocity of particle transport in flagella, they could, in principle, also drive some components of the slow transport system. For example, if ChAT-containing particles spent only a small fraction of time interacting with kinesin-II motors, this could lead to slow transport of ChAT. Alternatively, if the overall balance of anterograde versus retrograde transport of ChAT-containing particles is small, then net transport rates of ChAT would be slow. Finally, we note that if either of these parameters can be regulated by the axon, then the actual rate of ChAT transport might also vary in different cell types or circumstances, thus accounting for the diversity in rates reported. In this context, recent work demonstrating that kinesin-I may have a role in movement of intermediate filament subunits in nonneural cells is consistent with this general view (Prahlad et al. 1998). At present the question of how slow transport is driven remains open, but the possibility that kinesin-II motors might play a role warrants further investigation.

We thank S. Chen for isolating the cDNA clones and sequencing, and Joe Marszalek for providing the KIF3A cDNA. We are grateful to P. Salvaterra, J.T. Littleton, and K. Zinsmaier for providing various antisera.

S. Perez was supported by a National Science Foundation postdoctoral fellowship. This project was supported by a grant from the National Institutes of Health to L.S.B. Goldstein. L.S.B. Goldstein and H. Steller are investigators of the Howard Hughes Medical Institute.