Reciprocal localization of Nod and kinesin fusion proteins indicates microtubule polarity in the Drosophila oocyte, epithelium, neuron and muscle

The asymmetric segregation of cytoplasmic determinants is a general mechanism for distinguishing fates of sister cells, and requires the polar distribution of these molecules prior to cytokinesis (Horvitz and Herskowitz, 1992; Rhyu and Knoblich, 1995). The anterior-posterior (A-P) axis of the early embryo of Drosophila melanogaster offers an extreme example of asymmetrically distributed determinants (St Johnston and Nüsslein-Volhard, 1992). The mRNA for the anterior determinant bicoid (bcd) is localized to the anterior pole of the egg, while the mRNA for the posterior determinant nanos (nos) is localized to the ‘polar plasm ‘at the posterior pole (Berleth et al., 1988; Driever and Nüsslein-Volhard, 1988a,b; Lehmann and Nüsslein-Volhard, 1991; Wang and Lehmann, 1991). In addition, molecules necessary for development of the future germline, such as germ cell-less (gcl) mRNA, Vasa (Vas) protein and Tudor (Tud) protein, are highly enriched in the polar plasm (Hay et al., 1988; Lasko and Ashburner, 1990; Jongens et al., 1992; Bardsley et al., 1993). The localization of these determinants to their respective poles requires the establishment of an A-P axis during oogenesis.

Genetic analysis of the ‘posterior group ‘of maternal-effect genes has begun to unravel some of the mechanisms involved in assembling the polar plasm at the posterior pole (for reviews see St Johnston, 1993; Rongo and Lehmann, 1996). The oskar (osk) gene plays a critical role in polar plasm assembly. osk mRNA and Staufen (Stau) protein are the first components of polar plasm to localize to the posterior pole; their localization occurs during stages 8 and 9 of oogenesis (Ephrussi et al., 1991; Kim-Ha et al., 1991; St Johnston et al., 1991). osk is essential for polar plasm assembly, and mislocalization of a chimeric osk mRNA to the anterior of the oocyte leads to a stau-independent assembly of polar plasm at the anterior pole (Lehmann and Nüsslein-Volhard, 1986; Ephrussi and Lehmann, 1992). Normal embryonic polarity therefore relies upon the efficient transport of osk mRNA to the posterior pole of the oocyte.

The A-P axis of the oocyte depends on at least two intercellular signalling events: one from the oocyte to the follicle cells and a second one, dependent on the first, from the follicle cells back to the oocyte. As a stage 1 egg chamber assembles in the germarium, the oocyte positions itself posterior to the nurse cells and induces a subset of the adjacent follicle cells to become posterior polar cells (González-Reyes and St Johnston, 1994; for reviews of oogenesis, see Mahowald and Kambysellis, 1980; Spradling, 1993). The signal from the oocyte requires the germline function of the genes gurken (grk), which encodes a TGF-α homolog, and cornichon (cni), while its reception requires the Drosophila EGF receptor (DER or torpedo [top]) in the follicle cells (Neuman-Silberberg and Schüpbach, 1993; González-Reyes et al., 1995; Roth et al., 1995). If the signal is disrupted by mutation of the above genes or by abnormal positioning of the oocyte in spindle-A, spindle-B, spindle-C or spindle-D (spn-A, spn-B, spn-C or spn-D) mutants, the posterior follicle cells adopt characteristics typical of anterior follicle cells (González-Reyes and St Johnston, 1994;

González-Reyes et al., 1995; Roth et al., 1995). Posterior polar cells in turn signal the oocyte to establish normal A-P polarity; this second signal requires the functions of Notch (N) and Delta (Dl) in the soma and the catalytic subunit of protein kinase A (PKA, encoded by DCO) in the germline (Ruohola et al., 1991; Lane and Kalderon, 1994). Perturbing the somatic signal either by transformation of the posterior follicle cells or by mutations in N, Dl or DCO causes mislocalization of osk mRNA and Stau protein to the center of the oocyte, and mislocalization of bcd mRNA to the posterior as well as the anterior pole (Ruohola et al., 1991; González-Reyes and St Johnston, 1994; Lane and Kalderon, 1994; González-Reyes et al., 1995; Roth et al., 1995).

How do the follicle cells direct the localization of mRNAs within the oocyte? One popular model has been that the signal from the soma triggers a reorganization of the microtubule cytoskeleton and allows formation of a microtubule array, which is polarized along the A-P axis. The evidence for a reorganization of the microtubule cytoskeleton is clear from immunofluorescence studies (Theurkauf et al., 1992, 1993). Shortly after the 16-cell cluster of the oocyte and its nurse cells forms, a microtubule organizing center (MTOC) can be detected in the oocyte, posterior to the nucleus. Microtubules emanate from this MTOC into the nurse cells and evidence from electron microscopy indicates that the centrioles from the nurse cells accumulate in the oocyte (Mahowald and Strassheim, 1970). The MTOC and the centrioles are detectable through stage 6 of oogenesis, after which they disappear and, intriguingly, the localization of bcd and osk mRNAs begins (Mahowald and Kambysellis, 1980; St Johnston et al., 1989; Ephrussi et al., 1991; Kim-Ha et al., 1991; Theurkauf et al., 1992).

Evidence for a polarized microtubule array after stage 7 comes from two sources. First, immunofluorescence indicates that after the early MTOC disappears, oocyte microtubules form an anterior-to-posterior gradient; the high concentration of microtubules at the anterior cortex of the oocyte suggests an anterior nucleation of microtubules (Theurkauf et al., 1992). The second piece of evidence comes from experiments using a kinesin:β-galactosidase fusion protein (Kin:βgal, encoded by the fusion transgene khc:lacZ) (Giniger et al., 1993; Clark et al., 1994). The fusion protein contains the motor domain of the plus-end-directed microtubule motor protein kinesin (Yang et al., 1989, 1990) and localizes to regions of microtubule plus ends when expressed in cells with known cytoskeleletal polarity. When expressed in the germline, Kin:βgal accumulates in the posterior of the oocyte during stages 8-9 of oogenesis, which strongly suggests microtubule plus ends point to the posterior when osk mRNA is being localized (Clark et al, 1994). Moreover, in spn-A, spn-B, spn-C, spn-D, grk, cni, DER, N, Dl or DCO mutants, Kin:βgal, like osk mRNA and Stau protein, localizes to the center of the oocyte (Clark et al., 1994; González-Reyes and St Johnston, 1994; Lane et al., 1994; González-Reyes et al., 1995; Roth et al., 1995). Since Kin:βgal localization is independent of osk and stau (Clark et al., 1994), a simple hypothesis is that the signal from the follicle cells is necessary to polarize the microtubule cytoskeleton of the oocyte, which in turn directs the localization of osk and bcd mRNAs.

While polarization of the oocyte microtubules is thought to be an output of the signal from the follicle cells, the nature of this polarity has been questioned. According to the above model, plus ends of microtubules should point toward the posterior of stage 9 oocytes. However, both the heavy chain of cytoplasmic dynein, a minus-end-directed motor, and the Glued protein, a componenet of the dynein regulatory complex dynactin, are also posteriorly localized at stage 9 (Li et al., 1994; McGrail et al., 1995). This paradox creates some uncertainty about the polarity of the oocyte microtubule cytoskeleton and suggests that either Kin:βgal or dynein is not an accurate reporter of microtubule polarity of the oocyte. The localization of an independent molecule, in particular to the anterior, might help to resolve this dilemma.

In this paper, we construct a fusion protein, which localizes to minus ends of microtubules in several cell types and to the anterior cortex of stage 8-10 oocytes. The chimera is based on the kinesin-related protein (KRP) encoded by the no distributive disjunction gene (nod) (Zhang et al., 1990). nod is required for the proper segregation of nonexchange chromosomes via the distributive system during female meiosis in Drosophila (Carpenter, 1973; Grell, 1976; Hawley et al., 1993). In nod mutants, nonexchange chromosomes, such as chromosome IV, undergo high rates of nondisjunction. The Nod protein is distributed along the arms of meiotic chromosomes during metaphase I (stages 13-14 of oogenesis), due to an 82 amino acid region near the C terminus which contains three HMG-type DNA-binding domains (Afshar et al., 1995a,b). The N-terminal 320 amino acids of Nod shares strong sequence similarity with the motor domains of kinesin and other members of the superfamily of KRPs, which suggests that Nod may function as a microtubule motor in meiosis (Zhang et al., 1990; Goldstein, 1993). Knowing whether Nod is indeed a motor and identifying its direction is important for understanding the mechanism by which it ensures disjunction of achiasmate chromosomes. To date, however, the lack of motility of purified Nod protein along microtubules has left these questions unanswered.

We have created a Nod:β-galactosidase fusion protein (Nod:βgal) by replacing the motor domain of Kin:βgal with the putative motor domain from the N terminus of Nod. We find that, in cells with known microtubule polarities, Nod:βgal is a minus-end reporter for microtubules. Nod:βgal localizes to apical cytoplasm in polarized epithelial cells, to the poles of mitotic spindles in dividing cells and to the posterior of stage 1-6 oocytes. In stage 8-10 oocytes, Nod:βgal accumulates along the anterior cortex of the oocyte. The anterior localization of Nod:βgal in these oocytes supports the conclusions of microtubule polarity derived from the posterior localization of Kin:βgal. Furthermore, the in vivo localization of Nod:βgal in cells with known microtubule polarities may have implications for direction of the Nod putative motor domain.

UAS-KZ³⁰/TM3 and UAS-KZ^31Y stocks were generous gifts of P. Kolodziej. Line KZ1071 was obtained in a screen performed by P. Kolodziej, D. Doherty, T. Jongens, I. E. C. and E. Giniger (unpublished). GAL4 lines 24B and elav-GAL4 were generously provided by L. Luo. Mutant stocks y cv swa³v f/FM3, v swa¹/FM3, swa⁶/FM7, cn exu¹bw/CyO and vas¹cn exu bw/CyO were obtained from the Bloomington stock center; alleles are described in Lindsley and Zimm (1992).

To construct nod:lacZ, we made use of a nod cDNA plasmid, generously provided by H. Epstein and R. S. Hawley. The cDNA insert began at nucleotide −8 relative to the translation start site and was contained (from 5′ to 3′) between the KpnI and EcoRI sites of BluescriptIISK (Stratagene). The KpnI site was destroyed and converted to an XbaI site by addition of a linker to create pXba-nod. With plasmid khc:lacZ (Giniger et al., 1993) as a template, we used the polymerase chain reaction (PCR) to amplify a region of the coiled-coil domain, from nucleotides 1320 to 1598 of khc (Yang et al., 1989), and added HindIII and BamHI sites on the 5′ and 3′ ends, respectively. Primers for the reaction were GGGAAGCTTAAGAACGTGGTCT-GCGTTAACG and CGGATCCGGACATATTGGCGAGAGCCG. The amplified fragment was cut with HindIII and BamHI and used to replace the corresponding fragment of pXba:nod, to create pnod:khc. Integrity of the amplified sequence was confirmed by dideoxy sequencing (USB, Sequenase 2.0). Finally, an XbaI-SfiI fragment of pnod:khc was used to replace the corresponding fragment of pkhc:lacZ, thereby creating pnod:lacZ. For enhancer trap lines, nod:lacZ was cloned as a NotI-KpnI fragment into the vector pEG117 (Giniger et al., 1993) to create the pNZ transgene; for UAS lines, the NotI-KpnI fragment was cloned into the vector pUAS_T (Brand and Perrimon, 1993) to create pUAS-NZ. The transgenes were introduced into a y w stock by P-element-mediated transformation (Rubin and Spradling, 1982).

y w; NZ37.3/CyO; +/+ virgin females were crossed to y w; +/CyO; Sb Δ2-3/TM6,Ubx males to generate y w; NZ37.3/Cyo; Sb Δ2-3/+ F₁ males. These males were then crossed in pair matings to y w virgin females. Red-eyed, Curly, non-Stubble F₂ animals represented trans-positions of the NZ transgene; individual F₂ males were taken and crossed to y w virgin females to generate NZ stocks. In a few cases, F₂ females of this class were isolated to obtain transpositions onto the X chromosome; line NZ137.3 is one such insertion. Ovaries from NZ lines were dissected and stained for X-gal as previously described (Clark et al., 1994).

Preparation of an antisense RNA probe for lacZ and modifications to the technique of Tautz and Pfeifle (1989) were as described previously (Clark et al., 1994).

Ovaries were dissected by hand or prepared in mass and fixed by the technique of Theurkauf et al. (1992). For visualization of microtubules, ovaries were incubated for 10 minutes in cold (−20°C) methanol. Embryos were fixed for 20 minutes in a 1:1 mixture of hep-tane and 4% formaldehyde (Ted Pella, Inc.) in PEMS (100 mM Pipes pH 6.8, 2 mM EGTA, 1 mM MgCl₂). In cases where preservation of microtubules was necessary (e.g. spindle staining), embryos were fixed either for 2 minutes in a 1:1 mixture of heptane:37% formalde-hyde (Fisher) or twice for 5 minutes each in 1:1 heptane:methanol. Antibodies used were a polyclonal anti-β-galactosidase antibody from Cappel (preabsorbed against either egg chambers or embryos as appropriate, and used at 1:5000), monoclonal antibody DM1A against α-tubulin (Sigma, used at 1:500) and monclonal antibody mAb22C10 (generous gift of S. Benzer; used at 1:250). Anti-β-galactosidase stainings were amplified with biotinylated goat anti-rabbit IgG and either streptavidin:Cy3 (Jackson Laboratories, each at 1:200) or Vectastain ABC (Vector). Other secondary antibodies used were DTAF-conjugated donkey anti-mouse IgG (Jackson labs, 1:200). Peroxidase stained samples were dehydrated and mounted in Permount (Fisher). Fluorescent samples were mounted in SlowFade (Molecular Probes) and confocal images were obtained on a BioRad MRC600 confocal imaging system. Figures were assembled using Adobe Photoshop 3.0.5 and Adobe Illustrator 6.0.

The nod gene encodes a protein of 666 amino acids (Fig. 1A), the first 320 of which bear 36% identity to the motor domain (or ‘head ‘) of the Drosophila kinesin heavy chain (Zhang et al., 1990). The in vivo localization of Kin:βgal (Giniger et al., 1993; Clark et al., 1994) suggested an assay system to examine the function of the Nod head. Specifically, we asked whether the Nod head could functionally substitute for the kinesin head in Kin:βgal and promote localization of a fusion protein. We constructed a nod:lacZ fusion gene, encoding the fusion protein Nod:βgal (see Fig. 1 and Materials and Methods). The Nod:βgal fusion protein was identical to Kin:βgal except that the 333 amino acid kinesin head had been replaced by the 320 amino acid Nod head. Expression of nod:lacZ was accomplished by two different methods. (1) NZ lines utilized the ‘enhancer trap ‘technique, in which tissue-specific expression of the transgene is mediated by enhancers near the site of insertion in the genome (O‘Kane and Gehring, 1987; Bier et al., 1989; Grossniklaus et al., 1989). (2) UAS-NZ lines made use of the two-component expression system of Brand and Perrimon (1993), which is based on tissue-specific expression of the yeast transcriptional activator GAL4.

To examine the behavior of the Nod:βgal fusion protein, we first analysed its subcellular distribution in epithelial cells. Epithelial cells have a microtubule cytoskeleton which is polarized along the longitudinal axis of the cell, with minus ends at the apical side and plus ends pointing toward the basal side (Bacallao et al., 1989). We used two different types of epithelial cells for this experiment: cells of the embryonic hindgut and the follicle cells of the ovary. In the hindgut, the apical membrane is toward the lumen of the gut, while in the egg chamber, the apical membrane points toward the oocyte (Mahowald and Kambysellis, 1980; Skaer, 1993). As shown in Fig. 2, Nod:βgal localized to apical cytoplasm when expressed in either the gut or the follicle cells, whereas Kin:βgal localized to basal cytoplasm in both tissues. Thus, in epithelial cells of both the gut and the ovary, Nod:βgal and Kin:βgal exhibited reciprocal subcellular distributions.

Because Nod:βgal localized to apical cytoplasm in epithelial cells, we suspected that it might be an in vivo reporter for the minus ends of microtubules in polarized cytoskeletons. To test this further, we examined the distribution of Nod:βgal relative to the mitotic spindles of dividing cells. Mitotic spindle micro-tubules are oriented with minus ends at the spindle poles and plus ends toward the chromosomes (Euteneuer and McIntosh, 1981; Euteneuer et al., 1982). We stained 3-to 7-hour-old embryos from two enhancer trap lines, NZ118.3 and NZ153.1, with antibodies to α-tubulin and β-galactosidase. Nod:βgal was strongly concentrated at the poles of metaphase and anaphase spindles (Fig. 3 and data not shown). Some Nod:βgal was also found along the spindle microtubules; this was not due simply to ‘bleedthrough ‘from the tubulin channel, as similar staining was also detected when mitotic cells were identified with the DNA dye OliGreen (data not shown). Thus in dividing cells, as well as in the more differentiated epithelial cells, Nod:βgal accumulated in regions of cytoplasm that contained foci for microtubule minus ends.

In cells with defined microtubule polarity, Nod:βgal distribution consistently followed the positions of minus ends of microtubules. We therefore used it as an indicator for microtubule minus ends in the oocyte. Eight out of fifty transgenic NZ lines screened by X-gal staining of ovaries produced Nod:βgal in the germline. Subcellular distribution of Nod:βgal was consistent among these lines, despite differences in the timing and level of expression.

Both X-gal staining (not shown) and immunofluorescence demonstrated that, in stage 1-6 egg chambers from transgenic line NZ143.2, Nod:βgal accumulated specifically in a single germline cell at the posterior of the chamber (Fig. 4A,B). The consistent posterior position of this cell indicated that it was the oocyte. By stage 6, Nod:βgal was concentrated against the posterior cortex. The distribution of Nod:βgal during these early stages thus mirrored the documented position of the MTOC for the nurse cell/oocyte cluster (Theurkauf et al., 1992, 1993).

By stage 8, however, Nod:βgal underwent a dramatic redistribution within the oocyte. The protein was no longer detectable at the posterior of the oocyte, but was instead concentrated in a peripheral ring along the anterior cortex of the oocyte (Fig. 4C). Within this ring, the region flanking the oocyte nucleus had the highest levels of Nod:βgal. The anterior localization sharpened during stage 9 and persisted through stage 10 (Fig. 4D and data not shown). Localization of Nod:βgal was in striking contrast to that of Kin:βgal, which, as had been previously shown, localized to the posterior of stage 8 and 9 oocytes (Clark et al., 1994; Fig. 4E).

Localization of Nod:βgal was dependent upon the integrity of the microtubule cytoskeleton. To induce microtubule depolymerization, we fed NZ143.2 flies colchicine at a concentration of 100 μg/ml. After 4 hours of feeding, germline microtubules were undetectable and anterior localization of Nod:βgal was observed in only 8.1% (n=86) of stage 8-9 oocytes, as compared to 90% (n=120) from untreated flies (Fig. 5).

Whole-mount in situ hybridization with an antisense RNA probe for lacZ demonstrated that neither the early localization of Nod:βgal to the oocyte nor the later localization to the anterior of the oocyte was due to localization of the nod:lacZ mRNA (data not shown). The anterior localization of Nod:βgal was also independent of the genes exuperantia (exu) and swallow (swa), both of which are required for anterior localization of bcd mRNA during these stages (data not shown; St Johnston et al., 1989).

Microtubules in axons have a uniform orientation, with plus ends pointing distally from the cell body (Heidemann et al., 1981). Dendrites, in contrast, have been shown in vertebrates to contain a mixture of microtubules, with either plus or minus ends oriented distal to the cell body (Baas et al., 1988; Burton, 1988; Sharp et al., 1995). Since Nod:βgal and Kin:βgal acted as in vivo reporters for microtubule minus and plus ends, respectively, we used them to assess the overall polarity of the dendritic microtubule cytoskeleton. We crossed flies bearing either a UAS-NZ transgene or a UAS-KZ transgene to flies containing the pan-neural GAL4 transgene elav-GAL4 (Luo et al., 1994). Embryos produced by these crosses were collected and stained with both anti-β-galactosidase and the monoclonal antibody 22C10, which labels the cell membranes of all PNS neurons (Zipursky et al., 1984).

Fig. 6 shows the result of this experiment for the lateral cluster of chordotonal neurons. As expected, Nod:βgal was effectively excluded from the axons, a result consistent with the idea that Nod:βgal is a minus-end reporter. Surprisingly, though, Nod:βgal strongly accumulated at the distal tips of the dendrites, with some of the fusion protein also detectable along the shafts of the dendrites and in the cell bodies (Fig. 6A-C). In contrast, however, Kin:βgal efficiently filled the axon bundle for the chordotonal neurons, but only poorly labeled the dendrites, with no accumulation at the dendritic termini (Fig. 6D-F; see also Giniger et al., 1993). Similar observations were obtained for external sensory neurons (data not shown). These results suggest that the overall polarity of dendritic microtubules in the chordotonal and external sensory neurons of Drosophila is opposite to that of axons.

In addition to their reciprocal distributions in the oocyte, epithelial cells and neurons, Nod:βgal and Kin:βgal localized in complementary manners in embryonic muscles. We crossed UAS-NZ and UAS-KZ lines to the pan-mesoderm GAL4 transgene 24B, in order to express Nod:βgal and Kin:βgal in the embryonic musculature. As shown in Fig. 7A, Nod:βgal was excluded from the muscle attachment regions, but was detectable at high levels in the interior of the muscle fibers. In contrast, Kin:βgal concentrated at muscle attachment sites, and was present at only low levels throughout the remainder of the fibers (Fig. 7B and unpublished observations, E. Giniger and Y. N. J.).

The results presented here, together with previous results (Giniger et al., 1993; Clark et al., 1994), demonstrate that the Nod:βgal and Kin:βgal fusion proteins are in vivo reporters for polarized microtubule cytoskeletons. The two proteins, which differ only in the identity of the motor domain, localize in a reciprocal manner in a variety of different cell types. Furthermore, in cells in which the polarity of the microtubule cytoskeleton is known, such as epithelial cells, mitotic cells and neurons, the localization of Nod:βgal and/or Kin:βgal accurately reflects the polarity of the microtubules.

The localization of Nod:βgal during oogenesis mirrors the dynamic nature of the microtubule cytoskeleton. Early in oogenesis, as a discrete egg chamber forms in region 3 of the germarium, both the MTOC and Nod:βgal are found in the oocyte (Theurkauf et al., 1992, 1993; and Fig. 2A). At stage 7, the MTOC breaks down and the microtubule cytoskeleton reorganizes, and by stage 8, Nod:βgal localizes to the anterior cortex of the oocyte. The localization of Nod:βgal to the anterior of stage 8 oocytes indicates that by this time, microtubule minus ends point predominantly to the anterior. This finding supports the models of Theurkauf et al. (1992) and Clark et al. (1994), by which microtubules of stage 8-9 oocytes are polarized along the A-P axis, with minus ends at the anterior and plus ends at the posterior.

Like Nod:βgal, cytoplasmic dynein heavy chain and Glued protein both localize to the oocyte in stage 1-6 egg chambers and to the apical side of follicle cells. In stage 9 oocytes, however, dynein and Glued resemble Kin:βgal in that they localize to the posterior pole (Li et al., 1994; McGrail et al., 1995). It is still unclear why dynein localizes to the posterior of stage 9 oocytes. Given that Nod:βgal does not concentrate at the posterior pole at stage 9, we find it unlikely that dynein localizes to the posterior by minus-end-directed transport along the majority of microtubules, although such a mechanism is sufficient to explain its early accumulation in the oocyte. It is also unlikely that posterior localization of dynein is achieved by attachment to a specific site on the cortex at the posterior, since dynein, like osk mRNA, Stau protein and Kin:βgal, mislocalizes to the center of the oocyte in N mutants (Li et al., 1994). If dynein localizes to the posterior by minus-end-directed motility along a special subset of oppositely oriented microtubules, the organization of these microtubules would also have to be altered in the cytoskeletons of N mutant oocytes.

It is worth noting that subcellular localization of dynein may be mediated not only by its motor activity, but also by activities involved in recycling or binding to specific cargo. Indeed, in vertebrates, cytoplasmic dynein has been found at kineto-chores in mitotic cells (Pfarr et al., 1990; Steuer et al., 1990). It may be that dynein and Glued are brought to the posterior by association with one or more endogenous plus-end-directed motors. Alternatively, the direction of dynein motility may be regulated in vivo, so that its localization to the posterior occurs as a result of its own plus-end-directed motor activity. There is to date no evidence to support either of these models, and understanding the posterior localization of the dynein/Glued complex will await further analysis of its function in oogenesis.

The localizations of Nod:βgal and Kin:βgal during oogenesis mimics the localizations of several endogenous mRNAs and proteins, including bcd, osk, Stau, orb and K10 (St Johnston et al., 1989; Ephrussi et al, 1991; Kim-Ha et al., 1991; St Johnston et al., 1991; Cheung et al., 1992; Lantz et al., 1992). A simple hypothesis for the different types of mRNA localization during oogenesis is that early accumulation of specific mRNAs in the oocyte and later localization to the anterior pole are mediated by minus-end-directed microtubule motors, whereas localization to the posterior pole is mediated by plus-end-directed motors (Pokrywka and Stephenson, 1991; Theurkauf et al., 1992, 1993; Clark et al., 1994). While such models remain speculative, they suggest several steps at which mRNA localization may be regulated: (1) association of an mRNA with the appropriate motor, (2) activation of the motor-mRNA complex and (3) polarization of the microtubule cytoskeleton to specify the direction of transport.

Although microtubule motors that function in mRNA transport have yet to be reported, there is considerable experimental evidence to support a role for microtubules in mRNA localization (for reviews see Wilhelm and Vale, 1993, St Johnston, 1995). Microtubule inhibitor treatments prevent localization of mRNAs in oocytes from both Drosophila and Xenopus (Yisraeli et al., 1990; Pokrywka and Stephenson, 1991, 1995; Theurkauf et al., 1992, 1993; Clark et al., 1994). Localized mRNAs in both systems have been found biochemically to associate with microtubules in vitro (Yisraeli et al., 1990; Pokrywka and Stephenson, 1994). Particles of bcd mRNA and Stau protein, of an Exu-GFP fusion protein, and of myelin basic protein mRNA have been found to colocalize with microtubules in vivo (Ainger et al., 1993; Ferrandon et al., 1994; Wang and Hazelrigg, 1994). Finally, in spn-A, spn-B, spn-C, spn-D, grk, cni, top, N, Dl and DCO mutants, the mislocalization of bcd mRNA to both poles, and of both Kin:βgal and osk mRNA to the center of the oocyte, indicate a strong correlation between microtubule polarity and localization of maternal mRNAs for the A-P axis. As efficient localization of osk mRNA requires cytoskeletal non-muscle tropomyosin II, however, it is likely that the actin cytoskeletal also plays a critical role in the localization of maternal RNAs (Erdelyi et al., 1995; Tetzlaff et al., 1996).

Meiosis in wild-type females arrests in the first metaphase; exchange chromosomes are tightly aligned at the metaphase plate, while nonexchange chromosomes, such as the small fourth chromosome, are slightly displaced toward the spindle poles. In nod mutant females, however, the nonexchange chromosomes move precociously and excessively toward the poles, and are often ejected from the spindle as it is forming (Theurkauf and Hawley, 1992). nod is therefore required to prevent premature poleward migration of nonexchange chromosomes, either by exerting an anti-poleward force on the chromosomes or by providing a resistance to the poleward forces.

Our results suggest that the head of Nod encodes a minus-end-directed motor. Nod:βgal and Kin:βgal differ only in their heads, and yet the two proteins exhibit complementary localizations in a variety of cell types. We have not excluded the possibility that Nod:βgal localization is mediated by an association with an endogenous minus-end-directed motor. However, such a motor would have to be present in all cell types in which we see localization of the fusion protein and it would also have to interact specifically with Nod:βgal but not with Kin:βgal. Nevertheless, the complexity of the in vivo system precludes any strong conclusions about the direction of Nod and resolving this question will await the results of in vitro studies with purified or recombinant Nod protein. It is also possible that cofactors and/or chromosomes influence the direction of the Nod head in its native complex, which would complicate results from studies of recombinant proteins.

Afshar et al. (1995b) have found that the Nod head alone binds to microtubules in vivo without any obvious concentration at plus or minus ends, which suggests that other regions of the protein are necessary for motor activity. Motility of Nod:βgal may be facilitated by the kinesin coiled-coil domain or by the tetrameric β-galactosidase moiety at the C terminus. As Nod lacks a coiled-coil domain, it must either function as a monomer, like the KRP KIF1B (Nangaku et al., 1994), or form multimers in vivo without the benefit of its own coiled-coil. Multimerization could be achieved by the binding of accessory factors to Nod or by the binding of multiple molecules of Nod to adjacent sites on DNA.

Could Nod function in meiosis as a minus-end-directed motor? The polarities of meiotic spindle microtubules, which in female Drosophila are organized by the chromosomes, and of interchromosomal arrays of microtubules, which run between the nonexchange chromosomes at metaphase I, are both unknown (Theurkauf and Hawley, 1992; Hawley et al., 1993). If either of these microtubules arrays were oriented with minus ends toward the metaphase plate, a minus-end-directed protein could supply an anti-poleward force. A minus-end-directed motor could also be part of a ‘sliding ratchet, ‘coupling nonexchange chromosomes to the plus ends of growing microtubules; in such a model, the anti-poleward force would come from polymerization of the spindle micro-tubules themselves (Desai and Mitchison, 1995). Models based on Nod providing a resistance to poleward forces are also consistent with a minus-end direction; such models require that Nod or other chromosomal kinesins be regulated to allow the progression of chromosomes to the pole upon release from the metaphase I arrest.

Finally, it is worth noting that the only KRPs which have thus far been shown to be minus-end-directed motors (by in vitro motility of bacterially expressed protein) are the members of the KAR3/Ncd/CHO2 subfamily, all of which bear the motor domain at the C-terminus (McDonald et al., 1990; Walker et al., 1990; Endow et al., 1994; Kuriyama et al., 1995). However, in vitro studies on chimeras of kinesin heavy chain and the minus-end-directed motor protein Ncd indicate that direction is intrinsic to the head domain, and is not dictated by the position of the head at the N terminus or C terminus (Stewart et al., 1993; Chandra et al., 1993). Our observations of reciprocal in vivo localizations of two proteins differing only in their heads are consistent with the idea that motor polarity is encoded within the head. The mechanism by which direction is specified for microtubule motors is still unknown. Some clues may come from the recently reported crystal structures of kinesin and Ncd heads (Kull et al., 1996; Sablin et al., 1996). It is possible that Nod and another potential minus-end-directed N-terminal kinesin, CENP-E (Thrower et al., 1995), may help shed additional light on this problem. Because Nod and CENP-E belong to kinesin subfamilies different from each other and from Ncd, determining their directions with certainty and comparing their structures with Ncd may eventually help reveal the structural basis for polarity of microtubule motors.

We are grateful to H. Epstein, R. S. Hawley, P. Kolodziej, L. Luo and L. Zipursky for reagents and stocks, to D. Doherty, L. Luo, T. Mitchison, P. O ‘Farrell, S. Younger and anonymous reviewers for valuable comments on the manuscripts, to E. Giniger for communicating unpublished results, and to members of the Jan laboratory for valuable discussions throughout the course of this work. I. C. was supported by an NSF Predoctoral Fellowship and an NIH training grant. L. Y. J. and Y. N. J. are investigators of the Howard Hughes Medical Institute.