unc-83 encodes a novel component of the nuclear envelope and is essential for proper nuclear migration

Many cells have mechanisms to control the position of the nucleus during development and growth. For example, nuclei may migrate before an asymmetric cell division, and several epithelial organs show highly polarized nuclear positioning (Leung et al., 1999; Tomlinson, 1985). Defects in nuclear migration may lead to human disease. Mutations in LIS1 cause type I lissencephaly, in which neural migration into the cortex is disrupted (Dobyns and Truwit, 1995; Reiner et al., 1993). LIS1 is the human homolog of Aspergillus NudF. Mutations in NudF disrupt nuclear positioning in filamentous fungi, suggesting that mutations in LIS1 may cause lissencephaly by blocking nuclear migration (Morris, 2000). In most cases studied, nuclear migration appears to involve the centrosome and various motor proteins associated with the microtubule cytoskeleton (Raff and Glover, 1989). The centrosome is closely associated with the nucleus and may provide the force that drives the nucleus to a particular place in the cell (Reinsch and Gonczy, 1998). Microtubules, in coordination with molecular motors and associated machinery, play essential roles in various nuclear migration events (Morris, 2000; Reinsch and Gonczy, 1998). In budding yeast, dynein, dynactin and kinesins function to position the nucleus in preparation for cell division (Morris, 2000). Nuclear migration in the Drosophila eye involves the Glued protein, a component of dynactin, and the Klarischst protein, which plays a role regulating microtubules during nuclear migration (Fan and Ready, 1997; Mosley-Bishop et al., 1999). Not all nuclear migration events are dependent on microtubules and associated motors. For example, in Arabidopsis root hairs, nuclear migration appears to be mediated primarily by the actin cytoskeleton (Chytilova et al., 2000).

To better understand the regulation of nuclear migration, we are studying three sets of nuclear migrations that occur during the development of Caenorhabditis elegans. A group of epithelial blast cells, called P cells, have nuclei on the right and left lateral sides of the newly hatched larva; these nuclei migrate ventrally to form a single row in the ventral cord (Fig. 1) (Sulston and Horvitz, 1977). The P cells later divide to produce motoneurons and epithelial cells, some of which form the vulva, the egg laying and mating organ of hermaphrodites. During embryogenesis, left and right groups of dorsal epithelial cells intercalate, and their nuclei migrate to the contralateral side of the embryo. These cells subsequently fuse, forming a syncytium called hyp7 (Fig. 1) (Sulston et al., 1983). A third set of nuclear migrations occurs in the embryonic intestine, when the nuclei in the left and right groups of cells in the intestinal primordium move toward the future apical surface, where the intestinal lumen forms (Fig. 1) (Leung et al., 1999).

Genetic screens have identified two C. elegans genes, unc-83 and unc-84, that function in these three sets of nuclear migrations. Mutations in these genes were originally identified because their failure in P-cell nuclear migration leads to uncoordinated movement (Unc) and vulval defects (Horvitz and Sulston, 1980; Sulston and Horvitz, 1981). unc-84 encodes a protein of unknown function with a C-terminal SUN domain (for Sad1p, UNC-84) that is conserved from fission yeast to humans (Malone et al., 1999). UNC-84::GFP localizes to the nuclear envelope of most somatic cells in C. elegans (Malone et al., 1999). In S. pombe, the homologous protein Sad1p localizes to the nuclear envelope and spindle pole body, where it plays a role in spindle architecture (Hagan and Yanmagida, 1995).

We have analyzed the nuclear migration defects of unc-83 mutants and extend the characterization of unc-84 mutants. We show that neither gene appears to be necessary for centrosomes to associate with nuclei. We cloned the unc-83 gene and found that it encodes a novel protein associated with the nuclear envelope and is expressed in cells undergoing nuclear migration. The localization of UNC-83 to the envelope is dependent on unc-84(+) function. Our data suggest that UNC-83 and UNC-84 interact through the SUN domain of UNC-84.

Bristol N2 (Brenner, 1974) was used as the wild-type C. elegans strain and as the parent of all mutant strains. Mutants were isolated after mutagenesis with ethyl methanesulfonate (EMS). unc-83(ku18) was isolated in a screen for suppressors of the multivulva (Muv) phenotype of activated let-60(n1046)/ras (Sundaram and Han, 1995; Wu and Han, 1994). The alleles e1408, e1409, n320, n368 and n370 were isolated in screens for egg-laying defective (Egl) mutants (Horvitz and Sulston, 1980; Trent et al., 1983). Alleles n1727, n1766, n1826, n1827, n1866, n1883, n1886, n2011, n2012, n2100, n2101 and n2104 were isolated in screens for suppressors of the Muv phenotype of lin-15(n765) (Clark et al., 1992), and unc-83(n1497) was found to be a suppressor of the Muv phenotype of lin-1(e1275) (a gift from S. Clark).

For mapping unc-83, the strain DA1090 (exp-2(sa26sd) dpy-11(e224)/unc-46(e177) sDf30 V; mnDp26 – kindly provided by Leon Avery, University of Texas Southwestern, Dallas, TX) was used to create exp-2(sa26), dpy-11(e224)/unc-83(e1408) heterozygotes from which Dpy non-Let recombinants were picked. One of the seven such recombinants segregated unc-83(e1408). In the process of mapping and cloning unc-83, we rescued dpy-11(e224) by injection of cosmid T04F11, placing unc-83 in the small physical region between T04F11 and F12F3, which contains exp-2 (Davis et al., 1999). Cosmids covering this region were injected with a sur-5::GFP plasmid as a marker (Yochem et al., 1998) to test for rescue by standard techniques (Mello et al., 1991).

Defects in P-cell migration were quantified by counting the Pn.p daughter cells of the P cells in the ventral cords of L2 hermaphrodites raised at either 15 or 25°C using a microscope equipped with Nomarski optics (Sulston and Horvitz, 1977). We scored 11 of the 12 Pn.p cells; we did not score P12.p, the most posterior of the Pn.p cells. Alternatively, P-cell migration was quantified by counting P-cell-derived GABAergic neurons using the oxIs12(unc-47::GFP) marker (McIntire et al., 1997), which was crossed into an unc-83(e1408) background. Defects in hyp7 nuclear migration were quantified by counting the number of hyp7 nuclei found abnormally, using Nomarski optics, in the dorsal cords of L1 or L2 hermaphrodites raised at 20°C. We scored all 15 hyp7 nuclei that would be present in the dorsal cord if all migrations failed (Sulston et al., 1983). Defects in nuclear migration events leading towards intestinal polarization were scored as previously described (Leung et al., 1999).

Subclones of W01A11 were made as follows and tested for the rescue of the hyp7 nuclear migration defect. pD19 and pD21 were made by excising a 12.2 kb XhoI or an 8.8 kb AflII fragment, respectively, and re-ligating the W01A11 backbone. pD22 was constructed by ligating a 10.5 kb fragment of W01A11 from a EagI, HpaI double digest into the EagI and EcoRV sites of pBS (Stratagene, La Jolla, CA).

Expressed sequence tags (ESTs) were identified from the predicted open reading frame using The Intronerator software (Kent and Zahler, 2000). ESTs were kindly provided by Yuji Kohara (National Institute of Genetics, Mishima, Japan). The sequences of the cDNAs represented by ESTs yk230e1 and yk629c11 were determined.

dsRNA for RNAi experiments was made from a PCR product template with overhanging T7 promoters using T7 RNApol and injected into N2 adult hermaphrodites (Fire et al., 1998). L2 worms were scored for hyp7 and P-cell nuclear migration defects as described above 43-45 hours after injection. The 5′ RNAi experiment was directed against yk230e1 sequence corresponding to exons 3 through 7. The 3′ RNAi experiment was directed against exons 8 through 16.

Part of UNC-83, from the ATG of the short transcript in exon 8 to the beginning of the predicted transmembrane domain, was amplified by Pfu polymerase (Stratagene, La Jolla, CA) with appropriate overhanging restriction sites and cloned in-frame into the BamHI and SmaI sites in pGEX-2T (Amersham Pharmacia Biotech, Piscataway, NJ) to create a construct encoding an UNC-83/GST fusion protein or into the BamHI and PstI sites of pMAL-c2 (New England Biolabs, Beverly, MA) to create a construct encoding an UNC-83/MBP fusion protein. Both fusion proteins were expressed in E. coli strain BL21 codon plus (Stratagene) and purified on either glutathione sepharose 4B beads (Amersham Pharmacia Biotech) or amylose resin (New England Biolabs), according to the protocol supplied by the manufacturers.

Monoclonal antibodies were raised in six-week-old BALB/c female mice (Jackson Laboratories, Bar Harbor, ME). Six mice were injected three times at three week intervals with approximately 100 μg of the UNC-83/MBP fusion protein. One week after the last boost, test bleeds were examined against the UNC-83/GST fusion protein. The spleen was removed from the best responder 3 days after a final boost. The B cells were fused to a FOX-NY mouse myeloma at the University of Colorado Cancer Center tissue culture core facility (Denver, CO). 72 hybridoma clones were found to recognize UNC-83/GST by ELISA and further screened by immunofluorescence. Subclones of nine positive hybridomas were screened by immunofluorescence, and line 1209D7 was chosen as the best hybridoma based on a high signal-to-background noise ratio. All monoclonal antibodies were collected as tissue culture supernatants in RPMI + 15% fetal bovine serum. UNC-83 was not recognized on western blots using these antibodies; against whole wild-type or unc-83 null C. elegans extracts, using 12 different monoclonal antibodies, western blots were blank (data not shown). Either levels of UNC-83 were too low to be detected in whole worm extracts, or the monoclonal antibodies failed to recognize denatured UNC-83 on a western blot. The antibody did recognize overexpressed UNC-83/GST fusion protein on a western blot.

The UNC-84::GFP line has bee described previously (Malone et al., 1999); it was detected with GFP polyclonal antibody (Clontech, San Francisco, CA). The mouse monoclonal antibody IFA was used to detect embryonic centrosomes (Leung et al., 1999; Pruss et al., 1981). The rabbit polyclonal antibody against the C. elegans lamin homolog (Liu et al., 2000) was kindly provided by Y. Gruenbaum (The Hebrew University, Jerusalem, Israel). The anti-lamin antibody was used at a 1/250 dilution in PBST. A JAM-1::GFP line was made by integrating pJS191 (Raich et al., 1999) into N2 lines creating kuIs47 and crossed into an unc-83(e1408) background to identify adherens junctions.

For immunofluorescence with the UNC-83 antibody, embryos were extruded from slightly starved hermaphrodites, permeabilized by the freeze-crack method, fixed for 10 minutes in –20°C methanol, air dried and blocked in PBST (phosphate-buffered saline + 0.1% Triton X-100) +5% dry milk (Miller and Shakes, 1995). The fixed specimens were stained as described (Miller and Shakes, 1995). Primary antibody 1209D7 was used undiluted. Cy3-conjugated goat anti-mouse IgG and Cy2-conjugated goat anti-rat or anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA) diluted 1/200 in PBST were used as secondary antibodies. DNA was visualized by a 5 minute stain in 100 ng/ml of 4,6-diamidino-2-phenylindole (DAPI; Sigma, St Louis, MO) in PBST.

Images were collected using an Axioplan2 microscope (Carl Zeiss, Thorton, NY) and a Hamamatsu C4742-95 CCD camera (Hamamatsu Photonics KK, Bridgewater, NJ). Images were deconvolved and analyzed using Openlab 2.0.7 (Improvision, Lexington, MA) software, and figures were compiled using Photoshop 6.0 (Adobe, San Jose, CA), or as described (Leung et al., 1999).

The conserved SUN domain of UNC-84 (the C-terminal 198 amino acids) was amplified by PCR with appropriate overhanging restriction sites, cloned into the BamHI and EcoRI sites of pGEX-2T to create UNC-84-SUN/GST, and purified on glutathione sepharose 4B beads as described above. Alternatively, pGEX-2T was used alone to express GST. Crude extracts containing soluble UNC-83/MBP fusion protein were prepared as above. [³⁵S]methionine-labeled UNC-83 was made using the TnT-coupled reticulocyte lysate system (Promega, Madison, WI). The UNC-83B transcript, as represented by the EST yk230e1, was used as the template. The pulldown assay was preformed in a crude E. coli BL21 extract in PBS + 10% glycerol as previously described (Melcher and Johnston, 1995). Relative amounts of [³⁵S]methionine-labeled UNC-83 were detected and quantitated using a Storm PhosphorImager and ImageQuant software (Amersham Pharmacia Biotech).

In addition to the original unc-83(e1408) mutation, 18 other unc-83 alleles have been collected from various screens; many were isolated as suppressors of the multivulva phenotype caused by mutations that abnormally activate the Ras pathway in vulval development (see Materials and Methods). For each of these 19 alleles, we examined nuclear migration for the P cells and the precursors of hyp7 cells; representative alleles were also examined for nuclear migration in the intestinal primordium. We also examined pronuclear migration in newly fertilized unc-83(e1408) embryos and found the migration to be completely normal when compared with wild-type (data not shown).

P cell nuclear migrations often failed in unc-83 mutants. Nuclear migration for the P cells was quantified by counting Pn.p cell nuclei in the ventral cords of second larval stage (L2) hermaphrodites using Nomarski optics. A strong temperature-sensitive phenotype was observed for all unc-83 alleles similar to that reported for unc-83(e1408) animals (Table 1) (Sulston and Horvitz, 1981). While 90% of P cell nuclei migrated to the proper ventral position (termed the ventral cord) at 15°C, only 30-50% of the nuclei migrated to the ventral cord at 25°C. As most of these alleles probably eliminate unc-83 function , it is unlikely that the unc-83 gene product is thermolabile. Rather, our results support the hypothesis of Sulston and Horvitz (Sulston and Horvitz, 1981) that the process of P cell nuclear migration is intrinsically temperature-sensitive in the absence of unc-83 function.

The final positions of the P-cell nuclei depend on migrations of the P cells in addition to nuclear migrations within the P cells. Sulston and Horvitz have reported that P cells generally die in unc-83 mutant animals if their nuclei fail to migrate to the ventral cord (Sulston and Horvitz, 1981). By contrast, mutations in unc-73 or let-502 disrupt P cell migration before nuclear migration. In these mutants, the P cells survive and develop in lateral positions. To determine if P cells can develop laterally in unc-83 mutants, we scored P cell nuclear migration using an unc-47::GFP fusion gene, which is normally expressed in the 19 GABAergic neurons of the ventral cord (McIntire et al., 1997). Of these 19 GFP-positive cells, 13 are derived from P cells. In the wild type, all 19 cells were present (see Fig. 1B; n=15 worms). However, in unc-83(e1408) hermaphrodites raised at 25°C, we observed only 12.3 (±1.8) UNC-47::GFP-positive cells (Fig. 1C; n=16). In contrast to unc-73 and let-502 mutants, in which UNC-47::GFP-expressing neurons survive and develop in a lateral position (A. Spencer, C. J. M., M. H., unpublished), we did not observe any lateral UNC-47::GFP marked neurons in unc-83(e1408) worms.

We also examined the embryonic nuclear migrations of the precursors of the hyp7 hypodermal syncytium. Most unc-83 alleles caused the hyp7 nuclei to accumulate in a central region along the dorsal cord, rather than undergo the normal migration to contralateral positions. We counted hyp7 nuclei in the dorsal cords of L1 hermaphrodites using Nomarski optics. As the hyp7 nuclear migration defect is not temperature sensitive (Sulston and Horvitz, 1981), we scored worms raised at 20°C (Table 1; Fig. 1D-F). Twelve unc-83 mutants were severely abnormal in hyp7 nuclear migrations, one displayed an intermediate phenotype, and in six mutants hyp7 nuclei migrated normally (Table 1).

Nuclear migration was scored in the intestinal primordia of unc-83(e1408) and unc-83(ku18) animals. Both mutants had strong migration defects as indicated by nuclei not at the midline (future apical region) of the primordium at both 25°C and 15°C (Table 2; Fig. 1H-I). Surprisingly, mutant embryos raised at 25°C had slightly less severe defects than embryos raised at 15°C, in contrast to the effect of temperature on P-cell nuclear migrations. In some strains, the defect appeared to get worse as development progressed from the pre-bean to the bean stage. Perhaps unc-83 is required for nuclear positioning near the midline after migration. The intestinal cells defective in nuclear migration did not show the wild-type polarized distribution of cytoplasmic materials such as yolk and lipid vesicles. Nonetheless, the intestine developed an apical lumen, which functioned in feeding during larval development.

Several models for directed nuclear migration require that centrosomes associate with nuclei (Malone et al., 1999; Morris, 2000; Raff, 1999). Thus, defects in nuclear migration could result from defects in the association of centrosomes with nuclei. In the wild-type intestinal primordium, centrosomes are closely juxtaposed to the intestinal nuclei. The centrosomes are initially adjacent and either anterior or posterior to nuclei. However, before or during nuclear migration, centrosomes in the majority of intestinal cells rotate 90° to face near the midline of the intestinal primordium (Leung et al., 1999). We examined the distribution of centrosomes in the intestinal primordia of unc-83(e1408) mutants and found that centrosomes appeared to be associated with nuclei in all of the intestinal cells examined. Moreover, the centrosomes appeared to reposition toward the midline (Table 3; Fig. 2A-B). Thus, centrosomal re-positioning does not require nuclear migration. We also examined the distribution of centrosomes of the hyp7 cells in unc-83 mutants and found that centrosomes were associated normally with migrating nuclei (Fig. 2C). These data suggest that UNC-83 is not required at the nuclear envelope for centrosome attachment.

The nuclear matrix, composed of lamin and associated proteins, has been proposed to play an important structural role in nuclear migration (Gruenbaum et al., 2000). We examined lamin localization by immunofluorescence during hyp7 nuclear migration in wild-type and unc-83(e1408) embryos. Lamin appeared to be localized normally in unc-83(e1408) embryos (Fig. 3), suggesting that unc-83 is not essential for maintaining the gross structure of the nuclear matrix during nuclear migration.

Microtubules and the associated motor dynein also play important roles in many nuclear migration events (Gonczy et al., 1999; Mosley-Bishop et al., 1999; Reinsch and Gonczy, 1998). We therefore examined the distribution of microtubules in the intestinal primordium of unc-83(e1408) embryos. In all cases (n>25), microtubules were arranged as previously described for wild type (data not shown) (Leung et al., 1999). Of course unc-83 could be disrupting microtubule dynamics in a more localized pattern that we were unable to detect in the small cells in which nuclear migration events take place. Unfortunately, we were unable to examine the behavior of dynein heavy chain in unc-83 mutant cells during nuclear migration events, owing to their small size and high levels of cytoplasmic staining. However, dynein heavy chain was observed to be properly localized at the nuclear envelope in unc-83(e1408) early embryos and adult gonad cells (data not shown).

To better understand the role of unc-83 in nuclear migration, we cloned unc-83. unc-83 was originally mapped on chromosome V near dpy-11 (Horvitz and Sulston, 1980). We further mapped unc-83 to a narrow region between exp-2 and dpy-11 (Fig. 4A; see Materials and Methods). Cosmids in this region were injected and tested for their abilities to rescue the nuclear migration defects of unc-83(e1408) mutants. Cosmid W01A11 completely rescued the unc-83(e1408) nuclear migration defect of embryonic hyp7 cells. Large deletions in W01A11 were constructed and rescue results implicated an 11 kb region of W01A11 (Fig. 4B). Finally, a 10 kb subclone of W01A11 containing a single predicted open reading frame (W01A11.3) was found to rescue the hyp7 nuclear migration defect (Fig. 4B). However, we were never able to rescue the P-cell nuclear migration defect, even after injecting both W01A11 and the upstream cosmid ZK40.

We have detected three UNC-83 transcripts, UNC-83A, UNC-83B and UNC-83C, which encode proteins predicted to contain 1042, 974 and 741 residues, respectively (Fig. 4C). These proteins share a common 741 residue C terminus. UNC-83A and UNC-83B were identified by ESTs using The Intronerator software (Kent and Zahler, 2000). The cDNA yk629c11 (Kohara, 1996) represents UNC-83A. This cDNA is likely to contain the entire open reading frame, as an in-frame stop codon was detected 5′ to the first ATG. The UNC-83B transcript is defined by cDNA yk230e1 (Kohara, 1996). The 5′ end of yk230e1 contains part of an SL1 trans-spliced leader sequence which is spliced to the 5′ end of a majority of C. elegans transcripts (Huang and Hirsh, 1989), suggesting that we have identified the 5′ end of UNC-83B. The 5′ end of the UNC-83C transcript was isolated using PCR from a mixed stage λgt11 library (Okkema and Fire, 1994), with primers against an SL1 sequence and exon 10. This PCR product represents the complete 5′ end because it includes an SL1 leader sequence. We do not know if other UNC-83 transcripts exist.

All three of the identified UNC-83 transcripts extend at their 5′ ends beyond the genomic region covered by the minimal rescuing fragment pD22 (Fig. 4). This observation suggests that there may be additional unc-83 transcripts that start 3′ of the start of unc-83C or that another promoter is used to initiate transcription, possibly in the intron between exons 7 and 8.

To confirm that the unc-83A,unc-83B and unc-83C transcripts are responsible for unc-83 gene function, we determined the sequences of the open reading frames of 19 unc-83 mutants. We identified molecular lesions in 16 unc-83 alleles. All of the identified molecular lesions are severe, introducing a premature stop codon, mutating a splice site or, in one case, deleting 138 base pairs, which puts the rest of the protein out-of-frame (Fig. 4D). We did not identify lesions in the alleles n368, n2011 or n1727, suggesting that these mutations lie in the promoter or introns of unc-83.

The 16 identified unc-83 molecular lesions fall into two classes based on the severity of their hyp7 cell nuclear migration defects (see Table 1). The unc-83 alleles that lead to normal hyp7 nuclear migration, n2012, n320, n1766, n1827 and ku18, have molecular lesions that disrupt the open reading frames of UNC-83A and UNC-83B, but not UNC-83C. The alleles that disrupt hyp7 nuclear migration, n1883, e1409, n1866, n1497, n1826, n370, n2104, n1886, n2101, e1408 and n2100, have molecular lesions predicted to disrupt all three UNC-83 open reading frames. This result suggests that the shorter UNC-83C transcript is sufficient for hyp7 nuclear migration. Furthermore, as all alleles disrupt P cell nuclear migration (see Table 1), and both alleles tested (e1408 and ku18) disrupt intestinal nuclear migration, we propose that at least one of the longer UNC-83A or UNC-83B transcripts is necessary for P-cell and intestinal nuclear migration.

To examine further the tissue specificities of the different transcripts, we performed two RNAi experiments. RNAi directed against exons 8 to 16, which should disrupt all UNC-83 products, caused strong nuclear migration defects of both P cells and embryonic hyp7 cells (Table 4). By contrast, RNAi directed against exons 3 to 7, predicted to disrupt the open reading frames of UNC-83A and UNC-83B, produced a strong P-cell nuclear migration defect but no hyp7 nuclear migration defect (Table 4).

No proteins of high similarity to UNC-83 have been identified in BLAST searches with UNC-83 sequences. In addition, no obvious homologs have been found in the complete or nearly complete genomes of yeast, Drosophila or humans. Thus, UNC-83 is not obviously conserved at a sequence level, although it might be conserved at a structural level or with a similarity too low to be detected with current search tools. The only possible motifs we found in the UNC-83 sequence are a predicted transmembrane domain five residues from the C terminus (residues 1019-1037 of UNC-83A) and a short stretch of possible coiled-coil structure (residues 786-818 of UNC-83A).

To determine the subcellular localization of UNC-83, we raised monoclonal antibodies against a maltose-binding protein fusion protein containing most of UNC-83 (from the initiator methionine of UNC-83C to just before the carboxyl predicted transmembrane domain). The immunostaining patterns described below were not present in putative null unc-83 mutants and thus are specific for the UNC-83 protein (Fig. 5A).

Antibodies against UNC-83 stained the nuclear envelopes of migratory nuclei and several additional nuclei (Fig. 5, Fig. 6). UNC-83 co-localized with two other proteins of the nuclear envelope, UNC-84 and lamin (Fig. 6) (Liu et al., 2000; Malone et al., 1999). UNC-83 has a more punctate staining pattern than Lamin. We do not know what this staining pattern correlates to, but it is unlikely to be nuclear pores since staining with a nuclear pore marker (mAb414; Covane, Richmond, CA) is even more punctate (data not shown). We were unable to co-stain with both UNC-83 and nuclear pores, because both antibodies were raised in mice. It also appears that UNC-83 is more punctate that UNC-84 (Fig. 6). This may simply be due to the fact that UNC-84 was detected as a GFP fusion protein that is likely to be overexpressed (Malone et al., 1999).

Unlike UNC-84 and lamin, which can be detected at the nuclear envelope in nearly all cells, UNC-83 was detected in only a subset of cells. UNC-83 was not detected at the nuclear envelope of migrating pronuclei. We first detected UNC-83 at the nuclear envelopes of migrating embryonic hyp7 nuclei (Fig. 5B), consistent with the onset of the first observable unc-83 mutant defect. Later, at the bean stage of embryonic development, UNC-83 was localized to the nuclear envelopes of hyp7 cells, P cells and intestinal cells (Fig. 5D-F). By the comma stage, UNC-83 was detected in these cells as well as in several additional cells in the pharynx. unc-83(ku18) mutants, which have nuclear migration defects in the P cells and the intestinal cells but not in hyp7 precursors, showed immunostaining for UNC-83 in the hyp7 nuclei but not in the P cell nuclei or intestinal nuclei (Fig. 5C). In late stages of wild-type embryos, and in larvae and adults, UNC-83 was detected on nuclei in a wide variety of cells, including several cells around the pharynx and in the uterus.

Several of the nuclear migration defects of unc-83 mutants appeared very similar to those described previously for unc-84 mutants, and at least some of the migration defects did not appear to be exacerbated in unc-83; unc-84 double mutants (Malone et al., 1999). These results, together with the observation that UNC-83 and UNC-84 colocalize, suggest that these proteins could function in the same pathway or complex to control nuclear migration. We therefore examined the localization of UNC-83 in a collection of unc-84 mutant animals (Fig. 7E). We found that UNC-83 failed to localize to the nuclear envelope in the null mutant unc-84(n369) and in class 2 and 3 unc-84 alleles n323, n371, sa61 and n399, which have missense mutations in the conserved C terminus SUN domain of UNC-84 (Fig. 7A-C) (Malone et al., 1999). By contrast, UNC-83 was detected at the nuclear envelope in class 4 unc-84 alleles n322 and e1411, which have missense mutations or small deletions in the amino terminus of UNC-84 (Fig. 7D) (Malone et al., 1999). We could not determine if UNC-83 protein was expressed at normal levels in unc-84 class 1-3 mutants because our reagents do not function on Western blots. We conclude that UNC-84, and particularly the SUN domain of UNC-84, is required for UNC-83 localization to the envelope of migrating nuclei.

To determine if UNC-84 recruits UNC-83 to the nuclear envelope through a direct interaction, we performed GST pulldown assays. The SUN domain of UNC-84 fused to GST (UNC-84-SUN/GST) was attached to glutathione sepharose beads and used as the bait. As a negative control, GST alone was attached to beads and used as bait. [³⁵S]methionine-labeled UNC-83 bound to the UNC-84-SUN/GST-coated beads, but not the GST-coated beads (Fig. 7F). Addition of cold UNC-83/MBP competed with the [³⁵S]methionine-labeled UNC-83 so that only 39% of the [³⁵S]methionine-labeled UNC-83 was pulled down with the same amount of UNC-84-SUN/GST-coated beads (Fig. 7F). UNC-83 also interacts with itself in the same system (data not shown).

Using a monoclonal antibody, we found that UNC-83 localized to the nuclear envelope of a number of different cell types. The nuclear envelope consists of an inner nuclear membrane connected through nuclear pores to an outer nuclear membrane, which is continuous with the endoplasmic reticulum (Gruenbaum et al., 2000). As UNC-83 colocalized with lamin and was not widely distributed throughout the endoplasmic reticulum, UNC-83 may be a component of the inner nuclear membrane. However, this hypothesis needs to be tested using electron microscopy.

UNC-83 first localized to the nuclear envelope of migrating hyp7 precursor nuclei. Later in development, UNC-83 was also detected in the endodermal cells. Both of these cell types have pronounced defects in nuclear migration in unc-83 mutants, indicating a role for UNC-83 protein. However, UNC-83 was also detected at later stages of development in a wider variety of cell types, including several cells in the pharynx of comma-stage embryos and a number of hypodermal and other cells throughout the larval and adult stages. We do not know if UNC-83 functions in these cells.

The complex temporal pattern of UNC-83 expression may be controlled by enhancer elements spread throughout the large genomic region at the 5′ end of the gene. Each of the three unc-83 transcripts has a 4-5 kb intron near its beginning (Fig. 2). Such introns are large compared with average C. elegans introns, and often contain complex regulatory regions and transcript-specific promoters (Blumenthal and Steward, 1997). Additionally, the large regulatory region may explain why we were unable to rescue the P cell nuclear migration phenotype. It is possible that the molecular lesions in the three unc-83 alleles we failed to identify after determining the sequences of the entire predicted open reading frame reside within non-coding regulatory regions. Alternatively, these mutations may reside in the open reading frame of additional, unidentified unc-83 transcripts. For example, the minimal rescuing fragment for the unc-83 hyp7 nuclear migration phenotype does not contain a complete copy of any of the three identified transcripts and therefore may encode a distinct transcript.

We propose that the different unc-83 transcripts are regulated in a tissue-specific manner. Specifically, we suggest that the shortest identified unc-83 transcript, unc-83C, functions in hyp7 cells, while the longer ones, unc-83A and unc-83B, function in other tissues, including P cells and endodermal cells. Three observations support this model: (1) the unc-83 molecular lesions that do not disrupt the nuclear migrations of hyp7 cells do not disrupt the shorter transcript (Fig. 4D); (2) RNAi directed against only the longer transcripts does not disrupt nuclear migration in the hyp7 cells, whereas RNAi directed against all three transcripts does disrupt hyp7 nuclear migration (Table 4); (3) we detected UNC-83 at the nuclear envelope of hyp7 cells in unc-83(ku18) embryos (Fig. 5), in which hyp7 nuclear migration is not disrupted, but not in P cells or endodermal cells, which are disrupted for nuclear migration.

The unc-83 P-cell and hyp7 precursor nuclear migration defects we observed were similar to those previously described for null mutants of unc-84 (Malone et al., 1999). In addition, the unc-83 defect during intestinal nuclear migration was similar to the defects we observed in unc-84(n369) (a class 1, null allele) and unc-84(e1410) (a class 3 allele disrupting the conserved SUN domain; data not shown) mutants. However, unc-84(e1174) (a class 4 allele affecting the amino terminal end of UNC-84) only slightly disrupted intestinal nuclear migration (data not shown). In contrast to P-cell nuclear migration, intestinal nuclear migration events were cold sensitive, suggesting that a second partially redundant pathway can control intestinal nuclear migration.

As the phenotypes of unc-83 and unc-84 mutants are similar, we looked for a more direct interaction and found that mutations affecting the conserved SUN domain of UNC-84 disrupted the localization of UNC-83 (Fig. 7). This is different from unc-83 mutations, which do not disrupt the localization of UNC-84::GFP (Malone et al., 1999). We therefore propose that the role of UNC-84 in nuclear migration is to recruit UNC-83 to the nuclear envelope. The cell specificity of UNC-83 localization to the nuclear envelope could be defined by the complex transcriptional regulation of unc-83. This role of UNC-84 could completely account for the nuclear migration defects observed in unc-84 mutations. Consistent with this hypothesis, unc-84; unc-83 double mutant animals have the same nuclear migration defect as animals with either mutation alone (Malone et al., 1999), and UNC-83 and UNC-84 colocalize at the nuclear envelope (see Fig. 6). It is likely that the SUN domain of UNC-84 recruits UNC-83 to the nuclear envelope by a direct physical interaction, as we showed that the SUN domain of UNC-84 interacts with UNC-83 in vitro (Fig. 7F).

What is the role of UNC-83 in nuclear migration? Defects in nuclear migration could potentially be caused by abnormalities in the association between centrosomes and nuclei, as centrosomes can be critical in nuclear migration (Raff and Glover, 1989; Reinsch and Gonczy, 1998). We previously proposed that UNC-84 and UNC-83 function in the nuclear membrane to attach centrosomes to the nucleus for nuclear migration (Malone et al., 1999). This model predicts that in an unc-83 null mutant, centrosomes would migrate across the cell, but without functional UNC-83, the nucleus would fail to follow. Such a phenotype has been observed in both C. elegans early embryos and Drosophila embryos where mutations in the dynein heavy chain disrupt the centrosome to nuclear envelope association (Gonczy et al., 1999; Robinson et al., 1999). However, we found that in both hyp7 precursors and intestinal cells of unc-83 mutant embryos, centrosomes remain associated with the nucleus despite a defect in nuclear migration (Fig. 2; Table 3). Therefore unc-83 is not required for the attachment of centrosomes to nuclei.

Another possibility is that UNC-84 and UNC-83 interact with the nuclear matrix to maintain nuclear structure or promote dynein localization to the nuclear envelope during nuclear migration. These models predict that the nuclear matrix or dynein localization in unc-83 and unc-84 mutants would be disrupted. However, the nuclear matrix, at least as determined by lamin immunolocalization, appeared normal in unc-83 mutant hyp7 cells (Fig. 3). We were unable to examine dynein in migrating nuclei, owing to high levels of cytoplasmic dynein. However it is unlikely that UNC-83 and UNC-84 are required for dynein localization to the nuclear envelope because knockout experiments with dynein cause early pronuclear migration defects not seen in unc-83 or unc-84 mutant embryos (Gonczy et al., 1999).

We propose a model inspired by a recent report about nuclear positioning in fission yeast. Tran et al. (Tran et al., 2001) propose that microtubules push the nucleus to the center of the cell by growing at their plus ends at the cell periphery. Interestingly they show that Sad1p, the S. pombe homolog of UNC-84, is concentrated at regions of the nuclear envelope in contact with microtubules as well as spindle pole bodies. This suggests a role for Sad1p, and a homologous role for UNC-84 and UNC-83, in the association of microtubules with the nuclear envelope. This model could explain the punctate appearance of UNC-83 on the nuclear envelope. We do note that tubulin staining, in at least embryonic intestinal cells, appears normal in unc-83 or unc-84 mutants, suggesting that unc-83 nuclear migration defects are not caused by a gross disruption of microtubules. However, mutations in unc-83 could be disrupting a localized association between microtubules and the nuclear envelope beyond the resolution of our assay. In this model, UNC-83 and UNC-84 would function as part of a bridge to transfer the forces required for nuclear migration through the nuclear envelope between the structural elements of the nucleus, including the lamina, to molecular motors of the cytoskeleton. Specifically, UNC-83 may function to connect microtubules to the nucleus independently of centrosomes.

We thank Gail Ackerman, Kurt Christensen and Erika Unzueta for raising the monoclonal antibody, which was generated with support from the NCI, University of Colorado Cancer Center Core Grant P30-CA46934. We thank Yossef Gruenbaum for the lamin antibody, Allan Coulson for cosmids, Yuji Kohara for ESTs, Scott Clark and Glen Herrmann for unc-83 alleles, Leon Avery and the CGC for mapping strains, and members of the Han laboratory for advice. D. A. S. was supported by NIH postdoctoral grant F32-HL10449. G. H. was supported by a postdoctoral fellowship from the Damon Runyon-Walter Winchell Foundation (DRG1561). This research was supported by NIH grants GM47869 to M. H. and GM24663 to H. R. H., and The Howard Hughes Medical Institute, of which M. H., H. R. H. and J. R. P. are Investigators.