The LIS1-related Protein NUDF of Aspergillus nidulans and Its Interaction Partner NUDE Bind Directly to Specific Subunits of Dynein and Dynactin and to α- and γ-Tubulin*

The NUDF protein of Aspergillus nidulans, which is required for nuclear migration through the fungal mycelium, closely resembles the LIS1 protein required for migration of neurons to the cerebral cortex in humans. Genetic experiments suggested that NUDF influences nuclear migration by affecting cytoplasmic dynein. NUDF interacts with another protein, NUDE, which also affects nuclear migration in A. nidulans. Interactions among LIS1, NUDE, dynein, and γ-tubulin have been demonstrated in animal cells. In this paper we examine the interactions of the A. nidulans NUDF and NUDE proteins with components of dynein, dynactin, and with α- and γ-tubulin. We show that NUDF binds directly to α- and γ-tubulin and to the first P-loop of the cytoplasmic dynein heavy chain, whereas NUDE binds directly to α- and γ-tubulin, to NUDK (actin-related protein 1), and to the NUDG dynein LC8 light chain. The data suggest a direct role for NUDF in regulation of the dynein heavy chain and an effect on other dynein/dynactin subunits via NUDE. The interactions between NUDE, NUDF, and γ-tubulin suggest that this protein may also be involved in the regulation of dynein function. Additive interactions between NUDE and dynein and dynactin subunits suggest that NUDE acts as a scaffolding factor between components.

During growth of Aspergillus nidulans and other filamentous fungi nuclei migrate into the germ tube and distribute evenly along the cell length (1)(2)(3). Analysis of mutations that affect nuclear distribution in these fungi has shown that microtubules and the cytoplasmic dynein and dynactin complexes are the main components of the nuclear distribution machinery (3)(4)(5)(6)(7)(8). Three additional proteins, NUDC, NUDF, and NUDE, whose functions are still not well understood, are also required for nuclear migration in A. nidulans. NUDC is a 22-kDa protein required to maintain the intracellular concentration of NUDF (9). NUDF (a homolog of yeast PacIp) is a homodimeric 49-kDa protein with seven WD40 repeats (10) and a short predicted N-terminal coiled-coil domain that participates in homodimer formation (11). A. nidulans NUDF colocalizes with the NUDA dynein heavy chain in comet-like structures at the plus ends of cytoplasmic microtubules in vivo (12,13). The location of NUDA and NUDF at the ends of microtubules may have functional significance because the deletion of either protein increases microtubule stability (13). NUDE (RO11 in Neurospora crassa) is a 70-kDa protein that was initially shown to be required for nuclear migration in N. crassa (3). It was subsequently identified in A. nidulans as a high copy suppressor of a temperature-sensitive nudF mutation (14). NUDE has a long, very basic N-terminal coiled-coil domain and a C-terminal domain rich in serine and threonine. In the yeast twohybrid assay NUDE interacts with NUDF via the coiled-coil domains of the molecules. GFP 1 -labeled NUDE, like dynein and NUDF, also associates with the plus ends of microtubules (14). Genetic experiments in Aspergillus indicate that NUDF and NUDE influence nuclear motility by affecting cytoplasmic dynein (10,15).
NUDC, NUDF, and NUDE are evolutionarily conserved proteins with close homologues in many higher eukaryotes, including humans, rodents, Drosophila, and Xenopus (16 -22). The homologue of NUDF in higher eukaryotes is LIS1, which is particularly interesting because it is associated with a human genetic disease of children. Haploinsufficiency of LIS1 causes a devastating brain malformation known as Miller-Dieker lissencephaly, a disease characterized by a smooth cerebral cortex, severe mental deficiency, epilepsy, and a short life span (23). Lissencephaly is believed to be the result of impaired migration of neurons from the paraventricular area of the brain, where they replicate, to the cerebral cortex during development. LIS1 and NUDF are 42% identical in amino acid sequence, and both appear to be homodimers (11,24). We have suggested that NUDF and LIS1 affect a common motility mechanism involving cytoplasmic dynein which underlies both nuclear migration in A. nidulans and neuronal migration in animals (25). Considerable evidence to support this idea has been accumulated recently. The interactions among NUDC, NUDF, and NUDE initially characterized in Aspergillus are conserved in higher eukaryotes. Like NUDF, LIS1 interacts with microtubules and affects microtubule dynamics (26). It also affects nuclear migration (27), colocalizes with dynein (20,28,29), and affects dynein function in vivo (20, 30 -33). Rat NUDC interacts with LIS1 in the yeast two-hybrid system and in GST pull-down and coimmunoprecipitation experiments (25). Mammalian homologues of NUDE similarly interact with LIS1 and with various components of dynein and dynactin in indirect protein interaction assays (20,28,29). Interestingly, NUDE localizes to the centrosome in mammalian cells. It also interacts with ␥-tubulin and other centrosomal proteins in the yeast two-hybrid system and in coimmunoprecipitation experiments (20,22,26,28,29,31). The protein interaction experiments are informative; but because they were mainly done in unpurified systems, one cannot conclude that the observed interactions are direct, as they could be mediated by intervening proteins. In the present paper we have examined the interactions of the A. nidulans NUDF and NUDE proteins with each other as well as with component proteins of dynein and dynactin and with ␣and ␥-tubulin using yeast two-hybrid assays, coimmunoprecipitations, and pull-down experiments. We have additionally studied the binding of purified proteins to each other to determine whether their interactions are direct or indirect. The data indicate that both NUDE and NUDF bind to ␣-tubulin and ␥-tubulin, but they bind differently to dynein and dynactin. NUDF binds directly to the dynein heavy chain, whereas the NUDE protein is associated directly with the NUDG dynein light chain and the NUDK actin-related protein of dynactin. We also show that more than one dynein or dynactin subunit can bind simultaneously to NUDE, arguing for a possible function of NUDE as a scaffolding factor between components. The interactions between NUDF and NUDE and ␥-tubulin, also observed in mammalian cells (22), raise the additional possibility that ␥-tubulin may influence dynein function.

EXPERIMENTAL PROCEDURES
A. nidulans Media-YG (5 g of yeast extract and 20 g of glucose/liter supplemented with 0.1% trace elements) was used as a complete liquid medium for A. nidulans strains (34). Minimal medium (6 g/liter NaNO 3 , 0.52 g/liter KCl, 0.52 g/liter MgSO 4 , 1.52 g/liter KH 2 PO 4 , 2% glucose, and 0.1% trace elements) was from Pontecorvo et al. (35). 2% agar was added to solidify media where appropriate. To support the growth of strains carrying pyrG mutant alleles, media were supplemented with 10 mM uridine and uracil.
Construction of Yeast Strains Carrying Gal4p Fusion Proteins-The full-length, shortened and mutated constructs of NUDE, NUDF, LIS1, and human NUDE for use in the yeast two-hybrid system were described previously (11,14). Full-length open reading frames of A. nidulans nudK (dynactin actin-related protein), nudG (dynein light chain), nudI (dynein intermediate chain), tubA (␣-tubulin), tubB (␣-tubulin), tubC (␤-tubulin), benA (␤-tubulin), mipA (␥-tubulin), N. crassa ro10 (putative dynactin-associated protein) (3), ro2 (putative dynactin subunit) (36), and ro3 (dynactin p150 Glued ) (37) were made by polymerase chain reaction. Some N. crassa dynactin subunits were used for twohybrid analysis because N. crassa and A. nidulans are closely related ascomycetes, and the A. nidulans genes have not yet been cloned. The primers used contained additional restriction sites for insertion into the yeast two-hybrid plasmids pGBKT7 and pGADT7 (Matchmaker 3 system, CLONTECH). The open reading frame of nudA (dynein heavy chain) was inserted into both plasmids as three fragments of approximately equal length. The first third of NUDA included amino acids 1-1675. The second part encoded amino acids 1676 -3162, and the third contained the C-terminal coding sequence (amino acids 3163-4344). An additional two-hybrid NUDA fragment was constructed spanning the first P-loop of the dynein heavy chain (amino acids 1676 -2138) according to Sasaki et al. (20).
Yeast Two-hybrid Analysis-The yeast two-hybrid strain PJ69-4A was used containing three different reporter constructs driven by different Gal4p-dependent promoters (GAL1-HIS3, GAL2-ADE2, and GAL7-lacZ) (38). After pairwise transformation of plasmids, interactions were tested on media containing different amounts of 3-aminotriazole (3AT, 1 mM and 2 mM) or on medium lacking adenine. As a positive control we used the previously described interaction between NUDE and NUDF (14). Growth intensities were analyzed after a 3-day incubation at 32°C.

Construction of A. nidulans Strains Carrying Tagged Versions of Dynein, Dynactin, and Tubulin
Genes-Construction of a high copy plasmid pAAFS expressing S-tagged NUDF under the control of the alcA promoter and its integration into the nudF mutant strain XX20 resulting in NUDF expression strain CA1[pAAFS] was described previously (11). Construction of strain XX21ϫ6-17 expressing the VSV-Gtagged version of NUDE is described (14). For construction of a VSV-G-NUDE strain that also carries the pyrG marker, strain XX21ϫ6-17 was plated on uracil and uridine containing complete medium with 100 mg/ml 5-fluorotic acid to select against the functional pyrG gene. The resulting strain ANBH1 was unable to grow without uracil but was able to express the VSV-G-tagged version of NUDE. Open reading frames of nudE, nudG, nudK, nudA 1676 -2138 , and mipA were polymerase chain reaction amplified with specific primers each of which also contained the sequence GACTACAAGGACGACGACGACAAG encoding the FLAG epitope amino acids DYKDDDDK. The protein C epitope sequence GAGGACCAGGTCGACCCCCGTCTCATCGACGGTAAG encoding the amino acids EDQVDPRLIDGK was added to the open reading frame of NUDG and NUDI at both the 5Ј-and 3Ј-ends. Polymerase chain reaction products were integrated behind the inducible alcA promoter of the high copy plasmid pMCB17 (39) by replacing the GFP coding sequence. The functionality of the tagged proteins was tested by introducing the plasmids encoding these proteins into their relevant mutant strains and assessing their ability to restore growth under restrictive conditions. Transformation was carried out as described (40). Transformants were selected on medium without uridine and uracil to select for the presence of the prototrophic marker pyr4. Genes were expressed in A. nidulans strains GR5 and the VSV-G-NUDE containing strain ANBH1 in medium containing methyl ethyl ketone to induce the alcA promoter to obtain increased amounts of protein for purification. The names and genotypes of the A. nidulans strains are given in Table I.
Expression of Tagged Proteins-For high expression of tagged proteins the strains were grown in a preincubation medium (0.5% yeast extract, 1% glycerol, 0.1% trace elements) for 12 h, harvested, washed twice with H 2 O, and shifted to induction medium (0.25% yeast extract, 0.075% glycerol, 0.1% trace elements) containing 0.5% methyl ethyl ketone to induce alcA promoter-driven protein expression. Strains were grown for an additional 16 -20 h, harvested, and ground with a mortar and pestle in liquid nitrogen. Proteins were extracted by vortexing 5 g of mycelial powder in a 50-ml tube containing 20 ml of buffer that contained a protease inhibitor mixture for fungal extracts (Sigma). Buffers were those indicated for the isolation of the various tagged proteins according to the manufacturer's manuals. After a 4°C centrifugation at 3,000 ϫ g for 3 min to remove cell wall debris, supernatants were clarified by recentrifugation at 15,000 ϫ g for 15 min at 4°C. The resulting crude protein extracts were adjusted to a protein concentration of 4 g/l using the Bradford assay.
Coimmunoprecipitation and Pull-down Experiments-For immunoprecipitations all steps were performed on ice or at 4°C. Expression of tagged proteins was induced with 2% glycerol as the sole carbon source, which gives wild-type expression levels. 10 ml of crude protein extract was incubated on a rocker with 30 -40 g of monoclonal, specific antibodies against ␣-, ␤-, or ␥-tubulin (Sigma) An antibody raised against human isotype I and II ␤-tubulin was used as negative control (Sigma). 0.1 ml of protein G-agarose was then added, allowed to incubate for an additional 1 h, and the protein G-agarose-bound proteins were collected by centrifugation at 2,000 ϫ g for 1 h. The protein G-agarose beads were washed six times with 10 volumes of buffer. Bound proteins were resuspended directly in Laemmli sample buffer. Samples were boiled for 3 min, subjected to electrophoresis on a 4 -20% gradient SDSpolyacrylamide gel (Bio-Rad), and transferred to a nitrocellulose membrane. After incubation with primary antibody and anti-mouse alkaline phosphatase-conjugated secondary antibody, the blots were developed colorimetrically. For coprecipitation experiments, 10 ml of crude protein extract containing a single tagged protein was incubated with 100 l of tag-specific agarose for 1 h (anti-protein C affinity matrix, Roche Molecular Biochemicals; anti-FLAG M2 affinity gel, Sigma; S-protein agarose, Novagen; anti-VSV-G-agarose conjugate, Sigma). The agarose was washed six times with 1 ml of buffer, and the bound proteins of a small agarose sample were analyzed by SDS-PAGE and Coomassie Blue staining. Agarose-protein complexes were added to 10 ml of crude protein extract containing a second tagged protein version. Incubation continued for an additional 1 h. Agarose-protein complexes were washed six times, and proteins were eluted under native conditions according to the various manufacturers' instructions. The eluate was used for SDS-PAGE and Western blot analysis as described above. Our criterion for purity was the appearance of a single band on a Coomassie Blue-stained gel. Antibodies against the tags were ordered from the same companies as the affinity matrices.
Pull-down experiments with purified tagged proteins were performed with 1-5 g of each protein. Tagged proteins were purified from crude protein extracts with buffer containing 0.5-2% Tween 20 as detergent. Washing steps were performed as described above except that the last three washing steps were performed with buffer without detergent.
Proteins were mixed in a total buffer volume of 200 l and incubated on a shaker for 1 h on ice. 30 l of affinity-agarose specific to one or the other protein was added, and incubation was continued for an additional 1 h. Native eluted proteins were tested in Western blot analyses. For electrophoresis under native conditions eluted proteins were mixed in a total volume of 30 l and after incubation directly loaded on an 8% gel (see below).
Protein Gel Electrophoresis under Native Conditions-Protein shift experiments were performed under native gel electrophoresis conditions (i.e. without SDS). The best separations of protein complexes were obtained with an 8% acrylamide/bisacrylamide gel (pH 7.8). A 5% gel (pH 6.8) was used as stacking gel. Proteins were electrophoresed for 4 -16 h depending on the anticipated sizes of the protein complexes. Protein shifts with the dynein light chain NUDG were analyzed by cutting protein bands out of the native gel. Gel fragments were boiled for 10 min in SDS-buffer and transferred to a SDS-polyacrylamide gel. Proteins were visualized either by silver staining or by Western blot analysis.

RESULTS
NUDF and NUDE Interact with Subunits of Dynein, Dynactin, and Tubulin in the Yeast Two-hybrid System-We have used the yeast two-hybrid assay system to study the interactions of NUDF and NUDE with various components of cytoplasmic dynein, dynactin, and microtubules. Because of its large size we were unable to test the whole NUDA dynein heavy chain directly in the two-hybrid system. We therefore tested constructs that encoded fragments of the dynein heavy chain, each encoding about one-third of the molecule and an additional smaller fragment containing the P-loop. Other interaction targets included the NUDI dynein intermediate chain, the NUDG LC8 light chain, the NUDK ARP1 actinrelated protein of dynactin, and ␣-, ␤-, and ␥-tubulins. We also tested three components of the system from the related ascomycete N. crassa: the RO3 p150 component of dynactin (37), the dynactin-associated protein RO2 (36), and RO10, whose function in nuclear migration is not known (3). Different re- SC/-Leu/-Trp/-His with 2 mM 3AT; SC/-Leu/-Trp/-adenine. The first three media select for the expression of the HIS3 reporter gene, and the fourth medium selects for the expression of the ADE2 reporter gene. Growth in the absence of histidine or adenine is expected to result from interactions between the proteins encoded by the plasmids. Growth on medium SD/-Leu/-Trp/-His is indicated by a yellow dot and indicates a weak interaction between the expressed proteins. Stronger interactions correspond to growth on medium with increasing amounts of 3AT or lacking adenine and are indicated by green, blue, and red dots, respectively. Note that fusions with the NUDE coiled-coil domains and LIS1 activate the HIS3 gene expression in the absence of any interactions. 1 mM 3AT, a competitive inhibitor of the HIS3 gene product, suppresses the resulting background growth. Interactions were tested in both directions. Growth results for only one direction are shown because the results in the other direction were essentially identical. Two-hybrid interactions confirmed by direct protein-protein interactions are labeled by a hole within the colored dots. Direct interaction tests were only performed with A. nidulans proteins and not with the LIS1 and human NUDE used in the two-hybrid system. Direct interaction experiments between A. nidulans NUDF and NUDE have been described earlier (14). porter constructs were tested to determine the relative strength of the interactions (38). All of the proteins were tested as both bait and activators (Fig. 1). As described previously, NUDF interacted strongly with NUDE and with the coiled-coil region of NUDE (14). It also interacted with the dynein heavy chain fragment containing the first P-loop even though no interaction was seen with the middle one-third fragment of the dynein heavy chain, which includes the P-loop. This middle third fragment, however, also failed to show a two-hybrid interaction with any of the other proteins shown in Fig. 1, suggesting that the reason for its lack of reaction with NUDF may have been because it was not correctly folded (data not shown). NUDF also exhibited a moderately strong interaction with the NUDI intermediate chain and with the TUBA and MIPA ␣and ␥-tubulin proteins but did not interact with the NUDG dynein LC8 light chain or with the NUDK ARP1 subunit of dynactin. To determine whether these interactions involved the coiled-coil region of NUDF, we used a mutant NUDF construct in which the seven leucines involved in the coiled-coil interaction were mutated to alanines to destroy the ability of this region to undergo a coiled-coil dimerization. 2 This mutant construct failed to interact with any of the above components of dynein or dynactin as shown previously for the failed interac-tion between NUDE and the mutant NUDF. 3 This suggests either that the coiled-coil region of NUDF was directly involved in these interactions or that it was required to maintain the structure of NUDF. We attempted to test LIS1, the human homologue of NUDF, for its ability to interact with these same proteins, but the LIS1 protein by itself induced significant Gal4p-dependent transcription. LIS1 showed an interaction above this background only with ␣and ␥-tubulin. The fulllength NUDE protein interacted strongly with itself in the two-hybrid system, as did the NUDE coiled-coil domain. NUDE also interacted with NUDF (14), with the NUDI dynein intermediate and NUDG LC8 light chains, with the ARP1 of dynactin, and with the ␣and ␥-tubulin proteins. The NUDE coiled-coil domain gave a somewhat different interaction pattern. Like the full-length protein it gave a strong signal with NUDE, NUDF, and the dynein LC8 and intermediate chains, but it did not interact with the NUDK (ARP1) protein of dynactin or with the tubulin proteins. A human homologue of NUDE, hNUDE, which has a similar conserved coiled-coil domain, exhibited interactions similar to those of the Aspergillus NUDE except that it failed to interact with the NUDK ARP1 protein of dynactin (20). Similarly, the coiled-coil domain of hNUDE interacted with the same proteins as the coiled-coil 2 C. Ahn, unpublished. 3 V. P. Efimov, unpublished.

FIG. 2. NUDE and NUDF interact with ␣-, ␤-, and ␥-tubulin in coimmuno-and coprecipitation experiments. Panel A, coprecipitation experiments with purified proteins. Proteins NUDF (NUDF-Prot.S), NUDE (NUDE-Flag), NUDK (NUDK (ARP1)-Flag), and NUDG (NUDG (CDLC)-Flag)
were isolated by tag affinity purification. No detergent was used in order to preserve protein complexes. Proteins were eluted with S-protein and FLAG-protein, respectively, separated by SDS-PAGE, and stained with Coomassie Blue (lanes a). The tagged proteins are marked by a star. The proteins were transferred to a membrane, and Western blot analyses were performed (lanes b). NUDF and NUDE complexes were isolated from strains CA1[pAAFS] (NUDF-Prot.S) and XX21ϫ6-17 (VSV-G-NUDE), whereas NUDK and NUDG coprecipitations were performed with strains ANBH2 and ANBH5, which carry the FLAG-tagged versions of these proteins and the VSV-G-NUDE. Antibodies against the VSV-G tag were used to detect a coprecipitation of NUDE. As a size control, a protein standard is given (lane M). Panel B, coimmunoprecipitation experiments. Crude protein extracts of strain CA1[pAAFS] and XX21ϫ6-17 were incubated without antibodies and with antibodies specific against ␣-(␣-AB), ␤-(␤-AB), or ␥-tubulin (␥-AB). As a negative control a monoclonal antibody against the human isotype I and II ␤-tubulins was used (unsp. ␤-AB), which is not functional in A. nidulans. Antibodies and associated proteins were pulled out with protein G-agarose. Crude extracts containing no antibody were incubated with the same amount of protein G-agarose (Prot.G). Bound proteins were separated sequentially by SDS-PAGE and detected with antibodies either to the protein S tag of NUDF or to the VSV-G tag of NUDE. Note that ␣-, ␤-, and ␥-tubulins are not separated in this electrophoresis system. As a loading control, the Coomassie-stained protein G band is shown. domain of Aspergillus NUDE but did so less strongly. The Neurospora RO2, RO3, and RO10 proteins did not interact with Aspergillus NUDF or NUDE, the human homologues or the coiled-coil constructs.

NUDF and NUDE Coimmunoprecipitate and Pull Down Subunits of Dynein, Dynactin, and Tubulin from Crude Protein
Extracts-We next asked whether in vitro protein-protein interactions would corroborate the two-hybrid interactions. We had shown previously that a tagged version of NUDF coimmunoprecipitated NUDF from crude extracts via the coiled-coil domain (11) and that NUDF coimmunoprecipitated with NUDE (14), thereby confirming these two-hybrid interactions. We now have prepared new constructs containing tagged versions of dynein, dynactin, and the tubulin subunits. They include two different tagged versions of dynein light chain (NUDG-FLAG and NUDG-protein C tags), the dynein intermediate chain (NUDI-protein C-tagged), the dynein heavy chain first P-loop (NUDA 1676 -2138 -FLAG-tagged), the ARP of dynactin (NUDK-FLAG-tagged), ␣-tubulin (TUBA-FLAG-tagged), ␥-tubulin (MIPA-FLAG-tagged), and NUDE (NUDE-FLAGtagged). These were placed under the control of the alcA inducible promoter, transformed into Aspergillus, expressed, and affinity purified under mild conditions and without detergents as described under "Experimental Procedures." These proteins and their putative binding partners were used in the following pull-down and coimmunoprecipitation experiments. NUDE was pulled down from crude protein extracts by the purified tagged NUDG LC8 dynein light chain and by purified tagged NUDK ARP1 protein bound to agarose beads ( Fig. 2A). Similarly, VSV-G-tagged NUDE pulled down both NUDG and NUDK as well as other not yet unidentified proteins (data not shown). Antibodies to ␣and ␥-tubulin coimmunoprecipitated both NUDE and NUDF from wild-type crude protein extracts (Fig. 2B). Purified NUDE (FLAG-tagged) and purified NUDF (S-tagged) also pulled out ␣and ␥-tubulins ( Fig. 2A). These experiments confirmed most of the interactions seen in our two-hybrid experiments. Only the NUDI dynein intermediate chain, which interacted with NUDE and NUDF in the twohybrid system, failed to exhibit a physical interaction with these proteins.
The Purified NUDF and NUDE Proteins Interact Directly with Specific Dynein, Dynactin, and Tubulin Subunits-The yeast two-hybrid, coimmunoprecipitation, and pull-down experiments, although for the most part consistent with each other, nevertheless left open the possibility that the observed interactions could be indirect, i.e. mediated by a third protein.
We therefore retested each of the interactions observed in the crude systems using purified proteins to determine whether the purified proteins would interact in the same way. Each purified protein produced a single band of appropriate molecular mass (Fig. 3A), except for NUDE, which purified as a split band with a molecular mass of about 74 kDa in which both components stained with antibody against the FLAG tag (Fig.  3A). Purified NUDE protein interacted directly with purified NUDG dynein light chain in coprecipitation experiments. NUDE also interacted directly with purified NUDK, ␣-tubulin, and ␥-tubulin in pull-down experiments (Fig. 3B, right; see also Fig. 5A). Similarly, purified NUDF pulled down the first P-loop of the dynein heavy chain and tubulin proteins (Fig. 4A). These results were retested in protein shift experiments performed with purified proteins under native electrophoresis conditions. Under these conditions when the purified NUDF and ␥-tubulin

FIG. 3. Purification and interactions among dynein, dynactin, and microtubule subunits of A. nidulans.
Panel A, SDS-PAGE of affinity-purified tagged proteins. Extraction buffers contained 0.5-2% Tween 20 to prevent binding of associated proteins. The purified proteins were characterized by SDS-PAGE and stained with Coomassie Blue. The identity of each purified protein is indicated above the lanes. Lane M shows marker proteins. Panel B, interaction of NUDE with NUDG (left). 1 g of purified NUDE was incubated with the same amount of purified NUDG protein. As controls, each protein was incubated by itself. Complexes were isolated by adding affinity agarose against the FLAG tag of NUDE (FLAG-agarose) or the protein C tag of NUDG (ProteinC-agarose). Proteins were eluted under native conditions according to the manufacturer's instruction manuals and analyzed by Western blot hybridization with antibodies against both tags in parallel. Right, native gel electrophoresis. 1 g of NUDE and NUDK (ARP1) or NUDE and NUDG (CDLC) were mixed and incubated for 1 h on ice (ϩ) or loaded directly (Ϫ) on an 8% acrylamide/bisacrylamide gel (pH 7.8) native separation gel. A 5% gel (pH 6.8) was used as stacking gel. After electrophoresis proteins were detected with silver staining. proteins were coincubated and subjected to electrophoresis, a new band appeared which contained NUDF and ␥-tubulin. A similar result was observed for the interaction of NUDF protein with ␣-tubulin under native conditions (Fig. 4B). Interestingly, in these native gel separations, purified NUDF did not migrate into the gel except when it was bound to another protein. A new shifted protein band also appeared when the purified NUDE was coincubated with NUDK or NUDG proteins and subjected to native gel electrophoresis (Fig. 3B,  right). In each of these native gel experiments we showed by Western blotting that NUDE or NUDF and the second component were represented in the new band. NUDE did not interact with the dynein intermediate chain or the first P-loop of dynein heavy chain (data not shown), nor did NUDF interact directly with the NUDI dynein intermediate chain, the NUDG dynein light chain, or the actin-related protein NUDK. The absence of interactions between these proteins suggests that the interactions we have observed are likely to be specific.
More Than One Dynein and/or Dynactin Component Can Bind to NUDE at the Same Time-The experiments described above show that several subunits of dynein, dynactin, and tubulin can bind directly to the NUDE protein. To determine whether these interactions are additive or competitive, we performed native electrophoresis experiments with combinations of NUDE and pairs of proteins that interact individually with NUDE. Incubation of NUDE with the NUDK actin-related protein plus either ␣or ␥-tubulin caused the formation of new electrophoretically supershifted bands (Fig. 5A). Similar results were obtained when ␣-tubulin and NUDG were tested for their ability to interact simultaneously with NUDE. The NUDG dynein light chain and the NUDK actin-related subunit of dynactin could also bind simultaneously to NUDE, as shown by the presence of all three proteins within the same supershifted complex (Fig. 5B). In contrast to this additive binding, incubation of NUDE protein with both ␣and ␥-tubulin did not produce a supershifted band, indicating that only one tubulin molecule at a time can bind to NUDE. Possibly the tubulin proteins compete for the same binding site (Fig. 5A). DISCUSSION We have analyzed the ability of the A. nidulans NUDF and NUDE proteins to interact with subunits of dynein, dynactin, and tubulins in a variety of systems, the yeast two-hybrid system, coimmunoprecipitation and pull-down experiments, and most importantly by assaying the ability of the purified proteins to interact directly with each other. NUDE and NUDF exhibited distinctly different interaction patterns. The NUDF protein interacted only with the part of the dynein heavy chain which spans the first ATP P-loop binding site and with ␣and ␥tubulins; however, it did not interact with the middle third of the heavy chain, which contains the P-loop. The interaction of the first third of NUDA with the NUDG dynein light chain and of the last third with ␣and ␥-tubulins (data not shown) indicated that these fragments were folded properly (confirming known data from mammalian systems, 41), but we have no such evidence to demonstrate that the middle fragment folded correctly. Incorrect folding of the middle fragment containing the P-loop could explain the different binding affinity of NUDF to this fragment versus the P-loop fragment. NUDE also interacted with ␣and ␥tubulin. In contrast to NUDF it did not bind to the dynein heavy chain but rather interacted with the NUDG dynein light chain and with the NUDK actin-related protein of dynactin. Most of the proteins that produced strong two-hybrid interactions also interacted in the coimmunopre- FIG. 4. NUDF interacts directly with NUDA, TUBA, and ␥-tubulin. Panel A, in vitro protein binding assays. Left, 1 g of purified NUDF was incubated with the same amount of a purified NUDA fragment spanning the first P-loop. As a control, each protein was incubated by itself. Complexes were isolated by adding affinity agarose against the FLAG tag of NUDA (NUDA (CDHC) Flag-agarose) or the protein S tag of NUDF (NUDF ProteinS-agarose). Proteins were eluted under native conditions and analyzed in Western blot hybridizations with antibodies against tags of NUDA and NUDF in parallel. Right, the identical experiment was performed with NUDF and the ␥-tubulin protein MIPA. The Western blot shown was performed only with the FLAG antibody against the tag of ␥-tubulin. Note that ␥-tubulin is able to bind with a low affinity to protein S-agarose by itself. Panel B, protein shift experiments. Proteins ␥-tubulin and NUDF (NUDFϩ␥-tubulin, left) or NUDF and ␣-tubulin (TUBA) (NUDFϩ␣-tub, right) were incubated in combination and alone. Electrophoresis was performed under native gel conditions. Proteins were either silver stained or blotted for Western analysis. Antibodies used were specific to the FLAG tag of ␣and ␥-tubulin or the protein S-tag of NUDF. Under the gel conditions NUDF did not migrate into the gel. Therefore Western analysis is shown only for the right two lanes of each Coomassie-stained gel. Note that ␥-tubulin forms a high molecular mass complex under native conditions. Dynein light chain, dynein intermediate chain, and the actin-related protein NUDK did not form complexes with NUDF in native gels (data not shown). cipitation and pull-down experiments and interacted directly as purified proteins. Proteins that did not interact in the twohybrid system (as NUDF with NUDG or NUDF with NUDK) failed to interact in the purified system. The interactions and lack of interactions between the purified proteins essentially verified the two-hybrid results. Only one protein, NUDI, the dynein intermediate chain, exhibited discrepant results in the two-hybrid and purified systems.
The loss of all two-hybrid interactions by a strain containing point mutations in the coiled-coil domain of NUDF showed that NUDF interacts with proteins other than itself via its coiledcoil region. In contrast to NUDF, NUDE interactions appear to be shared between the coiled-coil and the C-terminal domains. Whereas NUDF and the dynein light chain bound to the Nterminal coiled-coil fragment of NUDE, the tubulin proteins and the actin-related protein NUDK did not, indicating that they bind elsewhere on the NUDE protein (Fig. 1). These data are consistent with our results with purified proteins from native protein shift experiments, which showed that NUDG and NUDK can bind simultaneously to NUDE. These experiments also argue for independent binding sites for NUDK and the tubulin proteins within the C terminus of NUDE. In contrast ␣and ␥-tubulin appear to bind to identical or overlapping regions. Many of our results were similar to observations made in mammalian systems. Higher eukaryotic homologues of NUDE also interact with LIS1 as well as with various components of dynein and dynactin and the centrosome in the yeast two-hybrid system and in copolymerization, coimmunoprecipitation, and other indirect protein interaction assays. These data from both mammalian and fungal systems suggest that NUDE functions as a bridge or assembly factor between tubulins and the dynein and dynactin complexes (20 -22, 28, 29, 33, 42). The experiments with mammalian NUDE and LIS1, however, leave open the possibility that the observed interactions could be mediated by intervening proteins. Our results extend these previous reports by showing that the interactions between NUDE, NUDF, and the various components of dynein, dynactin, and tubulin occur in the absence of any proteins other than those being tested. Thus these interactions are direct (Fig.  1). For the most part the NUDE and NUDF interactions are conserved between lower and higher eukaryotes. We did not, however, observe a two-hybrid interaction between NUDE and the dynein heavy chain, as reported in the animal system (20), suggesting that not all interactions are conserved between Aspergillus and higher eukaryotic dynein/dynactin-related proteins.
The interactions between NUDE, NUDF, and the various components of dynein and dynactin coincide with the locations of these proteins in the cell. We have shown by GFP labeling that NUDF, NUDE, and the cytoplasmic dynein heavy chain (NUDA) appear as comet-like structures at the plus ends of cytoplasmic microtubules in Aspergillus (12)(13)(14). We have determined recently that much of the GFP-tagged NUDA, NUDF, and NUDE protein is associated with membranous vesicles in Aspergillus, suggesting that the comets represent collections of vesicles. 4 We have also shown that a mutation in NUDK (the dynactin ARP1 protein) prevents the association of GFP-labeled NUDA with microtubules (12). These data suggest that dynactin is required to anchor dynein to the vesicles. In animal cells dynein can be seen at the plus ends of cytoplasmic microtubules under certain conditions (43). Although dynein and dynactin have been shown to be associated with endoplasmic vesicles (as well as other membranous structures) in animal cells (44), whether the dynein associated with microtubule ends is in vesicles is not known.
One of the most intriguing features of NUDF and NUDE, seen in both Aspergillus and animal systems, is their interac-4 B. Hoffmann and N. R. Morris, manuscript in preparation.
FIG. 5. Simultaneous binding of subunits of dynein, dynactin, and microtubules to NUDE. Panel A, in vitro protein shifts. Native electrophoresis experiments were performed with a mixture of three purified proteins. Protein mixtures were incubated for 1 h on ice (ϩ) or immediately loaded on the gel (Ϫ). Electrophoresis was performed for 16 h at 150V. Separated proteins were detected by silver staining and analyzed by Western blot hybridizations. ␣and ␥-tubulin were detected by Western blotting with specific monoclonal antibodies against these proteins. Panel B, Western blot analysis of supershift complexes containing NUDG. Purified proteins NUDE-FLAG, NUDG-protein C, and TUBA-FLAG (left) or NUDE-FLAG, NUDG-protein C, and NUDK-FLAG (right) were mixed and after incubation electrophoretically separated in a native gel for 10 h. The most shifted protein bands were cut out and boiled in 30 l of SDS loading dye for 10 min. Proteins were separated on a SDS-gel and blotted on a membrane. Proteins were analyzed with antibodies to the protein C tag and the FLAG tag. Note that the unbound NUDG protein needed ϳ3 h to move across the whole gel. E, NUDE-FLAG; K, NUDK-FLAG; G, NUDG-protein C; ␥, MIPA-FLAG; ␣, TUBA-FLAG. tion with ␥-tubulin. In animal cells NUDF and NUDE interact with ␥-tubulin and colocalize with it at the centrosome (20,22). Moreover, overexpression of NUDE causes ␥-tubulin to become dissociated from centrosomes, suggesting that there is a functional interaction between these proteins (22). Although we have no direct functional information to connect NUDE and NUDF with the spindle pole body, the centrosome equivalent, or with ␥-tubulin in Aspergillus, a recent observation made in Schizosaccharomyces pombe suggests a possible link to NUDE, NUDF, and dynein. An unusual cold-sensitive ␥-tubulin mutation (SL1) identified during a search for genes synthetically lethal with Pkl1p kinesin causes a mitotic block with an apparent impairment of chromosome attachment to kinetochore microtubules at low temperatures (45). It also causes excessive polymerization of microtubules. These phenocopy the effects of dynein deficiency seen in other organisms. Dynein deficiency has been shown to affect the interaction between spindle microtubules and kinetochores in Drosophila (46,47) and is well known to cause microtubule hyperpolymerization in Aspergillus and other fungi (7,12,13,48). Synthetic lethality between proteins often means that they share a common essential function. The fact that mutations in the dynein heavy chain are synthetically lethal with a number of different mitotic kinesins in Saccharomyces cerevisiae (49) suggested an explanation for the synthetic lethality of ␥-tubulin with Pkl1p. If an interaction of ␥-tubulin with NUDF and/or NUDE were required for proper dynein function, the synthetic lethality of ␥-tubulin with Pkl1p could represent an overlap between an essential function of dynein and Pkl1p kinesin similar to that seen in S. cerevisiae. In Aspergillus NUDE and NUDF interact with ␥-tubulin, but they do not colocalize with ␥-tubulin as in higher eukaryotic cells. NUDE, NUDF, and the dynein heavy chain are abundant at the plus ends of microtubules but are not found at the spindle pole bodies (the fungal equivalent of the centrosome) (12,13). Conversely, ␥-tubulin is concentrated at the spindle pole bodies but is not found at the plus ends of microtubules (50). It could be argued that the lack of colocalization of NUDF, NUDE, and dynein with ␥-tubulin in Aspergillus provides evidence against an interaction between ␥-tubulin and dynein in vivo. However, a significant fraction of ␥-tubulin is present in the cytoplasm in both Aspergillus 5 and in animal cells (51). The evidence that NUDE and NUDF interact physically with ␥-tubulin both in Aspergillus and in mammalian cells needs to be followed up with a functional explanation for this interaction. Whether ␥-tubulin affects dynein function and whether this is mediated by NUDE and/or NUDF need to be tested by direct experimentation.