Functional analysis of the trypanosomal AAA protein TbVCP with trans-dominant ATP hydrolysis mutants.

TbVCP is a member of the AAA (ATPases Associated with a variety of cellular Activities) family of proteins containing two ATPase domains. Southern analysis indicates TbVCP to have a single-locus, two-copy, genomic organization. One copy, but not both, can be disrupted by targeted gene replacement, suggesting that TbVCP is essential for trypanosome viability. Site-directed mutagenesis of the ATP hydrolysis motifs indicates that the second conserved ATPase domain is essential for TbVCP activity. Constitutive overexpression of TbVCP with a single mutation in the second hydrolysis motif or with mutations in both hydrolysis motifs was not possible. Regulated overexpression of these mutants resulted in cell death as a dominant negative phenotype. In each case cell growth arrested at 24-h post-induction and at all stages of the cell cycle as judged by replication of nuclear and kinetoplast genomes. Onset of growth arrest coincided with the development of severe and characteristic morphological alterations for each mutant. Neither constitutive nor regulated overexpression of wild type TbVCP or the single first hydrolysis domain mutant had any overt effect on cell viability or morphology. However, the distinct phenotype of the double mutant indicates that the first hydrolysis domain, although not essential, does modulate overall TbVCP function. Finally, yeast complementation studies demonstrated that TbVCP can functionally replace the yeast homologue Cdc48p, indicating that protein.protein interactions essential to function have been maintained over great phylogenetic distances.

Other AAA family members having two conserved ATPase domains include Pex1p and mammalian NSF (Sec18 in yeast) (15). Pex1p is essential for peroxisomal biogenesis in S. cerevisiae (16). NSF plays an essential role in the fusion cycle of transport vesicles with target membranes in many steps of eukaryotic secretory pathways, including endoplasmic reticulum to Golgi transport, intra-Golgi transport, and synaptic fusion (17)(18)(19). During the process of vesicle and target membrane fusion, NSF forms a 20 S complex with both SNAPs and membrane-bound SNAREs derived from both the vesicle and target membranes (20). Subsequent ATP hydrolysis by NSF is believed to mediate conformational changes leading to disassembly of SNARE complexes in post-fusion membranes (19).
The precise functions of VCP are less certain, but given its sequence and structural relationship to NSF, it is believed to play an analogous role in organellar biogenesis via homotypic membrane fusion (8 -10). This view is supported by evidence for interactions between VCP homologues and resident SNAREs of the endoplasmic reticulum (21,22) and Golgi (23). In mammalian cells these interactions are mediated by a soluble adapter protein called p47, which functions analogously to ␣SNAP in the formation of NSF 20 S complexes (24). In addition, a complex of cytosolic proteins involved in ubiquitin-mediated proteasome function (Ufd1) and nuclear import (Npl4) have been shown to physically interact with the VCP homologue p97 (25). The growing view is that VCP is a multifunctional protein that affects different activities via specific adapter proteins (25). The commonality of these functions may be the ATP-dependent protein folding/unfolding activity of VCP as has been demonstrated with the archaeal homologue (26).
Evidence for VCP function in cell cycle control stems from work with S. cerevisiae cdc48-1 mutants that arrest as large budded cells with an undivided nucleus in the bud junction (4). Because yeast undergoes closed mitosis, Cdc48p must enter the nucleus to affect its cell cycle-related function(s) and consequently contains a nuclear localization signal (27). Interest-ingly then, AtCdc48p but not porcine VCP can complement S. cerevisiae cdc48 mutants despite the fact that both plants and animals have open mitosis (6,27). However, VCP failed to complement even when deliberately targeted to the yeast nucleus (27) suggesting that VCP is unable to interact appropriately with yeast Cdc48p accessory proteins across the fungi: vertebrate phylogenetic divide.
The role of ATP hydrolysis in AAA protein function has been addressed by site-specific mutagenesis of individual ATPase domains. Mutations of the Walker A and B motifs in NSF indicate that the first ATPase domain (D1) is responsible for the bulk of the NEM-sensitive ATPase activity and that it is essential for NSF activity in in vitro fusion assays (28,29). Inactivation of the second domain (D2) had a modest effect on the total ATPase and fusion activities. In contrast, when this same strategy was employed with Pex1p, which functions in peroxisome biogenesis, it was found that the second ATPase domain, but not the first, was essential for activity in vivo (30).
To date no such analysis has been performed with any VCP homologue.
African trypanosomes are the causative agents of sleeping sickness in humans and nagana in cattle. Our laboratory studies the cell biology of these parasites focusing on secretory protein trafficking. In this work, we further our previous studies (1) by performing an in vivo functional analysis of TbVCP. Using a targeted gene disruption strategy we demonstrate that TbVCP is an essential gene for trypanosome viability. The nucleotide hydrolysis motifs of the two ATPase domains are inactivated by site-directed mutagenesis, and regulated expression of these mutants in transgenic trypanosomes is then used to assess the relative contribution of each domain to in vivo TbVCP activity. Our results indicate that the D2 domain is essential for TbVCP function and that the D1 domain, although not itself essential, must augment the D2 domain. Finally, yeast complementation experiments are performed to establish that TbVCP can functionally replace S. cerevisiae Cdc48p. These studies provide novel structure/function information on TbVCP as well as insights into its in vivo functions.

EXPERIMENTAL PROCEDURES
Trypanosome Growth, Nucleic Acid Preparation, and Immunological Procedures-Trypanosoma brucei strain 427 procyclic trypanosomes were utilized in targeted gene disruption experiments and for constitutive expression of TbVCP ATP hydrolysis mutants using the pKO vector series or the pXS2 neo vector, respectively (see below). Strain 29-13 procyclic trypanosomes (a strain 427 derivative) was used for regulated expression of TbVCP mutants. Strain 29-13 (genotype: TETR, HYG, T7RNAP, NEO) stably expresses the Tet repressor under hygromycin selection and the T7 RNA polymerase under neomycin selection. Transfection of this strain with reporter proteins in the pLew100 vector (see below) allows tetracycline-inducible expression from a procyclin promoter containing the Tet operator sequence (31). Strain 427 procyclic trypanosomes were maintained in TMP medium at 27°C as previously reported (32). pXS2 and pKO transformants of this strain were selected with 50 g/ml G418 and/or 25 g/ml hygromycin. Strain 29-13 was cultured at 27°C in Cunningham's medium (33) with 15% inactivated fetal bovine serum, penicillin/streptomycin, 15 g/ml G418, and 50 g/ml hygromycin. pLew100 transformants of strain 29-13 were selected with 2.5 g/ml phleomycin. Trypanosomal DNA was prepared according to the method of Medina-Acosta et al. (34). As needed, clonal cell lines were established by limiting dilution in conditioned media (1:1 fresh and spent culture media supplemented to 20% fetal calf serum). Metabolic radiolabeling, immunoprecipitation, and electrophoresis were as previously described (32).
pKO Vector Construction and Targeted TbVCP Gene Disruption-A 2410-bp EcoRI/XbaI fragment from pXS2 (35) containing 5Ј-3Ј: 1) the procyclin EP1-EP2 intergenic region, 2) the neomycin phosphotransferase gene (NEO), and 3) the tubulin ␤␣ intergenic region was cloned into the corresponding sites in pBluescript IISKϪ (pBSIISKϪ, Stratagene, San Diego CA) to generate pKO neo . Next the NEO cassette was replaced with a hygromycin phosphotransferase gene cassette (HYG) using flanking AscI/PacI sites (a generous gift of Dr. Ian Manger, Stanford University) to create pKO hyg . Specific TbVCP disruption constructs were assembled by first inserting a 349-bp HindIII/NheI 5Јflanking TbVCP genomic fragment (nucleotides Ϫ353 to Ϫ10 relative to nucleotide ϩ1 of the TbVCP open reading frame) into corresponding sites in the upstream polylinker portion of the pKO vectors. A 455-bp fragment was then amplified (primers JB165/166) from the 3Ј-flanking TbVCP genomic region and inserted into XbaI/SacI sites in the downstream polylinker portion of the pKO vectors. Digestion of these constructs (pKO neo ⌬vcp and pKO hyg ⌬vcp) with HindIII/SacI releases linear fragments bearing distinct drug selectable markers flanked by homology sequences for complete targeted replacement of the genomic TbVCP alleles.
Southern Analysis-ϳ3 g of genomic DNA was separated electrophoretically on 0.8% TAE-agarose gels and then capillary-transferred to nitrocellulose membranes (Nitropure, Micron Separations, Inc., Westboro, MA). Radiolabeled probe was generated by random priming (High Prime DNA labeling kit, Roche Molecular Biochemicals, Indianapolis, IN) utilizing [␣-32 P]dCTP (PerkinElmer Life Sciences, Boston, MA) according to the manufacturer's instructions. For determination of wild-type genomic organization, the 305-bp pVCP insert (1) was used as probe template; and for verification of the TbVCP single null disruptions, either the TbVCP 5Ј-UTR, or the NEO or HYG coding regions were used as probe template. Membranes were prehybridized (QuikHyb, Stratagene, San Diego, CA) for 30 Ϫ 60 min at 68°C followed by a 2-h hybridization at 68°C. High stringency final wash was 0.1ϫ SSC, 0.1% SDS, 60°C. Hybridization was visualized by autoradiography.
DExx Box Mutant Construction-The HincII-KpnI region of the pB-SIISK-polylinker was first deleted by digestion and blunt-end religation. A 2459-bp MluI/HincII fragment containing the entire TbVCP gene (83 bp of 5Ј-UTR, 2343 bp of open reading frame, 33 bp of 3Ј-UTR) was then blunt-end-cloned from the genomic clone p5VCP (1) into the EcoRV site of pBSIISK (⌬HincII-KpnI) to generate pTbVCP. The XhoI-KpnI fragment of pTbVCP was next cloned into XhoI/KpnI-digested pBSIISKϪ to generate pTbVCPXK.
Primary and secondary PCR reactions for the generation of DExx box (DB) ATPase hydrolysis mutants were performed according to a previous study (36) utilizing pfu DNA polymerase (Stratagene). Briefly, overlapping primary PCR products were amplified from p5VCP in two separate reactions. Each reaction utilized a unique flanking primer and a complementary mutagenic primer to introduce a single altered codon (E 3 Q) in the overlapping ends of the paired primary products. Secondary reactions utilized 10 ng each of the primary PCR products as templates with the two unique flanking primers. In the secondary reactions, denatured primary products re-anneal at the overlapping ends that contain the mutated codon and are amplified as one contiguous product. This process was performed independently for the two TbVCP ATPase domains, and each final product was sequenced though the region of PCR amplification to verify that only the desired mutation was generated.
For the DB1 mutant, primary PCR reactions were performed utilizing primer combinations JB149/168 and JB167/171; JB167 and JB168 encode the mutated nucleotide. Secondary reactions were performed with JB149 and JB171 as flanking primers. The secondary PCR product was digested with AgeI and XhoI, and the 348-bp internal fragment was inserted into AgeI/XhoI-digested pTbVCP to generate pTbVCP(E294Q).
Constitutive and Inducible Expression of TbVCP ATP Hydrolysis Mutants-The insertion of wild-type TbVCP into the constitutive expression vector pXS2 neo has been described previously (1). The entire DB1, DB2, and DB1/2 mutant TbVCP coding regions were excised from their respective vectors with HindIII/EcoRI and directionally cloned into the corresponding sites of pXS2 neo . Likewise, the wild type and mutant coding regions were cloned into the BamHI/HindIII sites of the inducible expression vector pLew100 (31). The resulting plasmids series are designated pXS2:WT, pXS2:DB1, etc., and pLew100:WT, pLew100: DB1, etc.
For inducible expression, strain 29-13 (NEO, HYG) trypanosomes were electroporated with 10 g of NotI-linearized pLew100:WT, pLew100:DB1, pLew100:DB2, or pLew100:DB1/2 by a modification of the standard protocol. After electroporation the cells were transferred to 10 ml of conditioned TMP media (1:1 fresh TMP and spent TMP supplemented to 20% fetal calf serum) containing G418 (15 g/ml) and hygromycin (50 g/ml). Transformants were selected on day 1 with phleomycin (2.5 g/ml). Viable transformants grew out in 2 weeks and were maintained initially in conditioned media, and then in Cunningham's medium. In this system homologous integration is targeted into the nontranscribed ribosomal RNA spacer, a transcriptionally silent region of the trypanosome genome (37). For induction of mutant TbVCP expression, tetracycline was added to log phase cultures at a concentration of 0.5 g/ml.
Microscopy-Wild type and mutant TbVCP cell cultures were seeded at 10 6 /ml and induced with tetracycline. Cells were harvested at 36 -40 h post-induction, washed, fixed, permeabilized, and stained with DAPI (1 g/ml) as described previously (1). Cells were viewed on a Zeiss Axioplan IIi at 100ϫ magnification. Differential interference contrast, and fluorescence images were collected digitally and merged using OpenLabs 2.2 software (Improvision Inc., Lexington, MA).
Yeast Complementation-The yeast expression vector pFL61 (2 plasmid) (38) containing the AtCDCc48 cDNA in the sense (pAtCdc48s) and antisense orientations (pAtCdc48␣s) (6) was kindly provided by Dr. Sebastian Bednarek (University of Wisconsin-Madison), as was the haploid S. cerevisiae strain MLY2006 (ts cdc48-2 ura3-52) (21). To generate sense (pTbVCPs) and antisense (pTbVCP␣s) constructs for expression in yeast the TbVCP-containing MluI/HincII genomic fragment was blunt-end-ligated into pFL61, and clones in both orientations were selected. Strains MLY2006 was transformed with each construct by the LiAc, polyethylene glycol transformation method (39). Ura ϩ transformants were selected on synthetic minimal media minus uracil at the permissive temperature (24°C). The Ura ϩ transformants were restreaked onto separate YPD plates and incubated at either the permissive or the non-permissive temperatures (37°C). To confirm the identity of the plasmids that complemented the conditional growth defects of MLY2006, the pTbVCPs construct was re-isolated according to a previous method (40). The isolated plasmid was verified by restriction enzyme mapping and subsequently retransformed into MLY2006 and challenged at the non-permissive temperature. In addition, positive transformants were restreaked onto synthetic minimal media containing 0.005% uracil (Sigma), 0.1% 5-fluoro-orotic acid (Sigma) to cause selective loss of the URA3-containing construct. Isolated colonies that grew on FOA were then restreaked onto YPD media and incubated at the non-permissive temperature.

TbVCP Is an Essential Gene-We wished to determine if
TbVCP is essential for trypanosome viability by a targeted gene disruption strategy. As trypanosomes are obligate diploids in which genes are often found in tandem chromosomal arrays (41,42), it was first necessary to define TbVCP genomic organization. The TbVCP genomic locus was mapped by restriction analysis (Fig. 1A), and representative single digests with endonucleases that cut once in and around the TbVCP  coding region are presented in Fig. 1B. Each enzyme produces a single hybridizing band of unique size (except ClaI and KpnI), suggesting that the TbVCP gene is not tandemly reiterated and most likely has two copies on homologous chromosomes. Detailed restriction analysis indicates that there are no gross polymorphisms in the two copies. 2 This interpretation of Tb-VCP genomic organization is supported by the results of gene knockout experiments (see below).
Knockout vectors (pKO neo ⌬vcp, pKO hyg ⌬vcp) were assembled containing NEO or HYG genes flanked by native sequences for robust RNA processing (procyclin and tubulin intergenic regions), which in turn were flanked by homologous targeting sequences for TbVCP locus (Fig. 1A). Electroporation of the excised HindIII/SacI fragment leads to targeted gene replacement of an entire TbVCP gene by homologous recombination. Expression of the selectable marker is mediated by endogenous read-through transcription.
Wild type and clonal cell lines transformed with either knockout construct were analyzed by genomic southern hybridization (Fig. 1C). Probing ClaI-and EcoRI-restricted wild type TbVCP/TbVCP DNA (TbVCP ϩ/ϩ ) with a TbVCP 5Ј-UTR probe revealed expected single fragments of ϳ5.7 and ϳ10 kb, respectively (ϩ/ϩ, lanes 1 and 2). This probe hybridized to two distinct fragments in restricted DNA from the TbVCP/⌬tbvcp:: neo (TbVCP ϩ/neo , lanes 3 and 4) and TbVCP/⌬tbvcp::hyg (TbVCP ϩ/hyg , lanes 5 and 6) clones. In each case, one fragment is the same size as wild type, and the second is either larger (ϳ6.0 or ϳ6.2 kb, ClaI, lanes 3 and 5) or smaller (ϳ6.5 kb, EcoRI, lanes 4 and 6). Each new fragment matches the predicted size for the desired homologous integration event consistent with the loss of a single TbVCP copy due to targeted gene replacement. Furthermore, the equal signal intensity of the wild type and integrant fragments support the presence of just two homologous copies of TbVCP in the wild type genome. To verify that the new bands were generated by targeted replacement of a wild type copy, the blot was sequentially stripped and hybridized with NEO-and HYG-specific probes. Again in each case, the probes hybridized with predicted fragments in the corresponding DNA only (TbVCP ϩ/neo , lanes 9 and 10; TbVCP ϩ/hyg , lanes 17 and 18).
We attempted to generate null TbVCP double knockouts by transformation of clonal TbVCP ϩ/hyg trypanosomes with pKO neo ⌬vcp; wild type TbVCP ϩ/ϩ cells were used as a positive control for transformation (Table I). All wild type recipients were selected with neomycin, and the TbVCP ϩ/hyg recipients were selected with either neomycin, or neomycin and hygromycin. As expected, viable neomycin-resistant transformants grew out of all wild type recipients, and Southern analysis confirmed that a single TbVCP copy was replaced. Viable transformants also grew out following selection of TbVCP ϩ/hyg recipients with neomycin alone. However, challenge of these transformants with hygromycin resulted in cell death, and Southern analysis indicated that the neomycin-selectable marker had replaced the hygromycin-selectable marker. 2 Simultaneous neomycin and hygromycin selection of TbVCP ϩ/hyg recipients resulted in no viable transformants. Collectively, these results establish that there are two homologous copies of TbVCP and that one copy, but not both, can be disrupted without effecting cell viability.
Constitutive Expression of TbVCP ATP Hydrolysis Mutants-Like other VCP homologues, TbVCP is an NEM-sensitive ATPase that has two active domains containing Walker A and B ATP binding and hydrolysis motifs (1). To assess the relative contribution of the two ATPase domains to total Tb-VCP activity, as well as the overall functional role of TbVCP in vivo, we generated specific ATP hydrolysis mutants by a strategy that has been previously used with mammalian NSF (28). Essential glutamic acid residues in each Walker B hydrolysis motif (also called the DExx Box) were mutated to glutamine creating the single ATPase mutants E294Q (DB1) and E567Q (DB2), and the double mutant E294Q/E567Q (DB1/2). Using the constitutive expression vector pXS2 neo (35), the wild type and mutant TbVCP genes were electroporated into both Tb-VCP ϩ/ϩ and TbVCP ϩ/hyg procyclic cell lines, and transformants were selected under drug pressure.
Viable transformants were consistently generated in both wild type TbVCP ϩ/ϩ and single null TbVCP ϩ/hyg -recipient trypanosomes electoporated with the wild type gene (Table II). Viable transformants were also consistently generated with the DB1 mutant in both recipient strains. These transformants grew out within the same period of drug selection (ϳ1 week) as those transformed with the wild type gene. However, viable transformants were rarely generated when wild type trypanosomes were electroporated with the DB2 or DB1/2 TbVCP mutant genes and were never generated in the single knockout recipients. The few wild type recipients (one for each construct) that did survive drug selection took 3-4 times longer to grow out, but then grew similarly to the wild type and DB1 transformants. A likely explanation for these neomycin-resistant cell lines is that some genetic rearrangement occurred that eliminated or inactivated the mutant gene without loss of the NEOselectable marker. Collectively, these results suggest that overexpression of TbVCP with a mutated D2 ATPase domain exerts a trans-dominant negative phenotype on endogenous wild type TbVCP.
Inducible Expression of TbVCP Hydrolysis Mutants-To examine the phenotypic effect of these mutations we exploited a system for regulated expression of toxic gene products in trypanosomes (31). The strain 29-13 procyclic host cell line for this system is engineered to constitutively express both T7 RNA polymerase and the Tet repressor. The polymerase is expressed under neomycin selection by endogenous readthrough transcription in the tubulin locus. Tet repressor expression is linked to hygromycin resistance and is driven by an attenuated T7 promoter (from within its integration target site in the RNP 1 locus). The inducible expression vector, pLew100, contains both a wild type T7 promoter that drives expression of a phleomycin resistance gene and a tetracycline-responsive procyclin promoter for regulated expression of toxic gene prod-

trans-dominant TbVCP ATP Hydryolysis Mutants
ucts. The wild type and mutant TbVCP genes were inserted into pLew100 and linear plasmids were transfected into 29-13 host cells. Clonal triple-resistant (neo, hyg, phleo) cell lines were selected and analyzed for growth kinetics and TbVCP synthesis under tetracycline induction (Fig. 2). All uninduced cell lines grew with normal kinetics, as did the wild type and DB1 clones in the presence of tetracycline (Fig. 2A). The DB2 and DB1/2 cell lines, however, ceased growth within 24 h of induction concomitant with characteristic alterations in cellular morphology by 36 -40 h (see below). Over the ensuing 2-3 days, these mutant cells became spherical with only the tip of the flagellum protruding 2 and by days 5 and 6 post-induction the cells died. All cell lines overexpressed TbVCP polypeptides with similar kinetics, and to the same extent, under tetracycline induction (Fig. 2B). Elevated expression was not seen in uninduced cells and quantitative enzyme-linked immunosorbent assays indicate that maximum induction (at 24 h) typically reaches 3-to 5-fold over levels of endogenous TbVCP. 2 These results are consistent with the failure to constitutively overexpress either the DB2 or DB1/2 mutant and support the conclusion that when expressed they exert a negative transdominant phenotype over endogenous TbVCP. To further investigate this issue we measured the stability of endogenous TbVCP in untransformed cells by pulse-chase radiolabeling. Essentially no turnover was observed over a 48-h period (Fig.  2C), well beyond the onset of growth arrest in the induced DB2 and DB1/2 cell lines. This confirms that the phenotypes are trans-dominant in that they are manifested in the continued presence of endogenous wild type protein.
Phenotypic Analysis of the TbVCP ATPase Mutants-We first assessed the effect of the DB1 and DB2 mutations on the ability of TbVCP to oligomerize. VCP homologues from other systems are known to be homo-hexamers, and our previous studies are consistent with a similar oligomeric structure for TbVCP (1). Overexpression of wild type and mutant proteins was induced in the 29-13 cell lines, and the oligomeric state of TbVCP in cytosolic extracts was determined by velocity sedimentation. Immunoblotting of gradient fractions (Fig. 3) indicates that most of the TbVCP in the wild type and mutant cell lines was assembled into oligomers of the correct size (fractions 9 -12). In each case a lesser amount of TbVCP polypeptide sedimented at a position intermediate to that expected for monomer and dimer. In no case, however, did the amount of non-hexameric TbVCP in mutant extracts differ significantly from that in the wild type extract, or were significant amounts of TbVCP polypeptide detected at the bottom of the gradients. These results suggest that there are no gross defects in the ability of the ATP hydrolysis mutants to properly oligomerize in vivo.
Trypanosomes have a characteristic morphology in which the flagellum emerges via the posterior flagellar pocket, adheres along the length of the cell body, and extends forward at the anterior end. The nucleus is centrally located and the kinetoplast (the genome of the single mitochondrion) is positioned near the base of the flagellum. Genome replication and cell division occurs in a clear morphologically defined order (43). Most cells in an asynchronously growing population are in interphase, which is characterized by a single flagellum, kinetoplast, and nucleus (1k/1n, ϳ85%). During cell division flagellar duplication initiates first, followed closely by kinetoplast

trans-dominant TbVCP ATP Hydryolysis Mutants
replication generating intermediates with two kinetoplasts and a single nucleus (2k/1n, Ͻ10%). Next nuclear division and flagellar repositioning occurs to produce cells with two complete flagella, kinetoplasts and nuclei (2k/2n, Ͻ10%). Ensuing longitudinal cytokinesis produces two interphase progeny. Thus, the approximate position within the cell cycle can be determined for any individual cell by assessing organellar genome copy number.
Using these criteria we have evaluated the effect of overexpression of TbVCP proteins on trypanosomal morphology and cell cycle progression at an early period (36 -40 h) when wild type cells were in log phase growth but the mutants had ceased cell division. Induction of wild type TbVCP altered neither the normal proportion (Fig. 4) nor the gross morphology of any stage of the cell cycle (Fig. 5, A-C). Likewise, the DB1 mutant remained normal in all aspects following induction. 2 Induction of the DB2 mutant, however, led to growth-arrested forms with a characteristic "spindle" morphology in which the cell body had apparently retracted forward leaving a prominent posterior spine (Fig. 5, D-I). In extreme cases the spine is completely devoid of cytosol as judged by immunofluorescent imaging for HSP70. 2 Arrested spindle cells were observed at all normal stages of the cell cycle including 1k/1n, 2k/1n, and 2k/2n (Fig. 5, D-H), but it was rare to find a cell arrested in the subsequent process of cytokinesis. Interestingly, the percent of interphase cells was reduced relative to wild type controls, and this was balanced by an increase in products of aberrant cell division (Fig. 4, 1k/1n versus other). The simplest of these were anucleate (1k/0n) cells (Fig. 5I) similar to the "zoids" generated by treatment with anti-microtubule agents (44), but other multinucleate and/or multikinetoplastic forms were also seen (not shown). Such forms are extremely rare in wild type cells.
The alteration of morphology with the double DB1/2 mutant was more extreme than the DB2 phenotype (Fig. 6). Again cells arrested in all normal stages of the cell cycle (Fig. 6, A-C), but with a characteristic "tadpole" morphology in which the cell body appeared retracted from the anterior end. Consequently, tadpoles had rounded posterior cell bodies containing the organellar genomes and narrow cytoplasm-containing anterior cell bodies with adherent flagella. Like DB2, the proportion of cells arrested in interphase was decreased relative to wild type, and the level of aberrant division products was elevated (Figs. 4 and 6, C-E). These included both products of uneven division, anucleate zoids (1k/0n) and the matching 1k/2n cells (Fig. 6D), as well as other multigenomic (Ͼ2k/Ͼ2n) cells (Fig. 6E). But unlike DB2, many cells with fully replicated organellar genomes also arrested in subsequent stages of cytokinesis (Fig. 6,  A and B). Most prominent in this category were cells that arrested following complete longitudinal fission. These cells presented as oppositely oriented daughters, each with complete flagella and bodies containing segregated organellar genomes but broadly fused across a transverse central plane (Fig. 6B).
Yeast Complementation-In a final analysis we performed complementation studies in S. cerevisiae to determine if Tb-VCP could functionally replace the yeast homologue, Cdc48p. The TbVCP genomic coding region was cloned into the yeast expression vector pFL61 (38) in the sense and antisense orientations. Controls for complementation included the pFL61 vector alone and the pFL61 vector with the Arabidopsis homologue AtCdc48 in both orientations. AtCdc48 has previously been demonstrated to complement both cold-and temperature-sensitive S. cerevisiae cdc48 conditional lethal mutants (6). The S. cerevisiae cdc48-2 temperature-sensitive strain MLY2006 was transformed with the pFL61 constructs, and Ura ϩ transformants were selected at the permissive temperature. Transformants were subsequently restreaked and challenged at the non-permissive temperature (Fig. 7B).
Neither untransformed MLY2006 (Fig.7, sector a), transformants carrying empty pFL61 vector (sector b), nor transformants carrying either antisense gene (sectors c and e) were able to grow at the non-permissive temperature. Both sense TbVCP (sector f) and sense AtCdc48 (sector d) transformants were able to grow at the non-permissive temperature, but the AtCdc48 colonies were smaller and took longer to develop than those Individual cells were examined by differential interference contrast and fluorescence microscopy as described under "Experimental Procedures"; digitally merged images are presented. The status of organellar genome replication is indicated for each cell; and the normal positions of kinetoplast (k) and nuclear (n) genomes are indicated for each wild type cell. All cells are presented with the anterior end oriented toward the top.

trans-dominant TbVCP ATP Hydryolysis Mutants
carrying TbVCP (longer incubation not shown). This finding is consistent with the observation that AtCdc48 can complement the cdc48-8 temperature-sensitive strain KFY189 at 36°C but not 37°C (6). Two experiments were performed to confirm the complementation analyses. First, the sense TbVCP plasmid was re-isolated from two complementing transformants, verified by restriction enzyme mapping, and then retransformed into MLY2006. The transformants again grew at the non-permissive temperature (sector g) verifying that the TbVCP gene is responsible for the complementation. Second, sense TbVCP transformants that were grown in FOA, which causes the loss of URA3-containing plasmids, lost the ability to grow at the non-permissive temperature (sector h). These results demonstrate that trypanosomal VCP is capable of functionally replacing the yeast homologue. DISCUSSION Using African trypanosomes as a model system we have investigated the in vivo functions of VCP/Cdc48p in eukaryotic cells. This protein is clearly essential for trypanosome viability as one, but not both, copies of the endogenous gene can be knocked out by targeted gene replacement. This is consistent with findings in S. cerevisiae where Cdc48p is essential for viability and growth of haploid cells (4). Disruption of one endogenous copy of TbVCP had no effect on viability or morphology despite a modest reduction of TbVCP protein as judged by immunoblot analyses. 2 Additional studies with single knockout TbVCP ϩ/hyg cells were performed to assess whether reduction of TbVCP copy number had subtle effects on secretory trafficking. Three reporter genes were introduced using the stable constitutive expression vector pXS2 neo (35). BiPN is a soluble secreted reporter; BIPN MDDL is retained in the endoplasmic reticulum; and 117wt is a glycosylphosphatidylinositol-anchored cell surface reporter (35,45). Deletion of a single TbVCP copy had no effect on the transport kinetics or localization of any of these reporters in the TbVCP ϩ/hyg genotype. 2 Like many members of the AAA family, e.g. NSF and Pex1p, TbVCP has two conserved ATPase domains. Specific mutagenesis of the nucleotide binding and hydrolysis motifs in these domains has been used to study the relative contribution of each to the overall function of both Pex1p (30) and NSF (28,29). These studies indicate that the Pex1p D2 domain and the NSF D1 domain are essential for in vivo and in vitro function of these proteins. Interestingly, sequence similarity between the essential domains of NSF and Pex1p is greater than the similarity between the two domains within each individual protein (16). In contrast, both ATPase domains of VCP/Cdc48p homologues are highly similar to each other, as well as with the essential domains of NSF and Pex1p. We have used the same mutagenesis approach to investigate the relative contribution of each domain to TbVCP function in vivo. A single critical glutamate residue within the DExx box sequence of each ATP hydrolysis motif was changed to glutamine to create the DB1 (D1 domain) and DB2 (D2 domain) single mutants, and the DB1/2 double mutant. The analogous mutations in NSF reduced NEM-sensitive ATP hydrolysis by 69% (D1), 28% (D2), and Ͼ90% (D1/D2) indicating that the activities of the individual domains are additive and that the combined mutations can completely ablate total ATPase activity (28). We have not directly assayed ATP hydrolysis in our TbVCP mutants, but considering the effects of such mutations on NSF and the dramatic phenotypic alterations demonstrated herein, it is likely that these mutations significantly reduce the hydrolytic potential of each ATPase domain.
When the TbVCP mutants were independently transfected into procyclic trypanosomes using a constitutive expression vector, no transformants were obtained with either construct containing the DB2 mutation, whereas wild type and DB1 transformants were readily generated. Given the low turnover rate of endogenous TbVCP (t1 ⁄2 Ͼ 48 h), this finding suggests that the DB2 and DB1/2 mutants exert a trans-dominant negative effect over endogenous wild type TbVCP, thereby preventing outgrowth of transgenic cells. To assess the phenotypes associated with the DB1 and DB1/2 mutants, we resorted to expression in a regulated transcription system (31). Induction of all constructs in clonal cell lines was rapid and equivalent, leading to a 3-to 5-fold overexpression relative to endogenous wild type TbVCP. As found with constitutive overexpression, neither wild type nor DB1 TbVCP had any effect on cell growth or morphology. Induction of the DB2 and DB1/2 mutants, however, resulted in complete cessation of cell growth within 24 h, much faster than the turnover of endogenous TbVCP confirming that these mutants exert a trans-dominant negative effect.
TbVCP is a homo-hexamer, and, if ATPase activity by one or both domains is essential for assembly, it is possible that overexpression of the mutants could inhibit oligomerization. This could have several potential outcomes. Mutant TbVCPs may fail to assemble leaving the endogenous wild type protein to form functional homo-hexamers, a scenario that could account for the lack of phenotype with the DB1 mutant. Alternatively, the mutant subunits may block assembly of all TbVCP proteins, or may be included in mixed hexamers with wild type subunits that are then dysfunctional in downstream events mediated by TbVCP. Either of these scenarios could account for the lethal phenotypes of the DB2 and DB1/2 mutants. However, when examined by velocity sedimentation no gross defects in oligomerization were observed with any form of TbVCP. Given the 3-to 5-fold overexpression of these proteins, it is likely that all hexamers contain subunits of both endogenous and induced TbVCP. Thus, although ATP hydrolysis by all subunits is clearly not required for oligomerization, hydrolysis by the 1-2 wild type subunits in each hexamer cannot be excluded as essential for the assembly process. Nevertheless, our results indicate that the trans-dominant effects of the mutants are likely exerted over some downstream function(s) of hexameric TbVCP.
Following the arrest of cell growth (ϳ24-h post-induction), both the DB2 and DB1/2 mutants assumed characteristically altered cell morphologies (DB2, spindles; DB1/2, tadpoles). Neither arrested in any particular stage of the cell cycle (1k/1n, etc.), but in each case there was a significant elevation in the number of cells with aberrant mitochondrial and nuclear genomes. Given the number of cellular processes in which other VCP homologues are known to participate, our findings cannot illuminate the mechanism(s) by which the TbVCP mutant phenotypes are affected in trypanosomes. However, several conclusions are possible. First, organellar genome profiles of arrested cells suggest that TbVCP is not a cell cycle regulator per se. Were this the case we would expect cells to arrest at a specific stage of organellar replication. The yeast gene, CDC48, was originally identified as a cell division cycle mutant (46), and this is one of the many functions typically assigned to its homologues. However, the unique features of trypanosome organellar replication allow a direct assessment of the cell cycle, and our results argue against a direct function in this process, at least in trypanosomes. Second, the elevated level of aberrant genomes must result from either unequal segregation of daughter organelles (1k/0n and 1k/2n progeny), from cytokinesis without organellar replication (1k/0n and 0k/1n progeny), or from ongoing organellar replication without cytokinesis (3k/3n, etc). Such aberrations are never observed in normal cells, suggesting that TbVCP may have a role, direct or indirect, in cytokinesis. Finally, the fact that the DB1/2 mutant has a subtle but distinct phenotype relative to DB2 alone suggests that there is a hierarchy of TbVCP functions in which D1 hydrolytic activity may be required. Such a conclusion would not be obvious from the apparent lack of phenotype when the DB1 mutant is expressed in isolation. Thus, although the D2 domain seems to be paramount in TbVCP activity, the D1 domain must also contribute to overall function (discussed below).
VCP/Cdc48p homologues have been implicated in many other cellular functions, the best characterized of which are proteasome activity (13,14,25) and homotypic membrane events, such as endoplasmic reticulum fusion (10,22) and Golgi fusion (8,9). In each case, the ATPase has been shown to physically interact with either soluble (13,14,24,25) or membrane-bound (21,22,23) cofactors and substrates for ATP-dependent folding/dissociation reactions. Molecular chaperone activity may in fact be the common feature of all VCP/Cdc48p functions, and there are at least two soluble accessory proteins that bind to and regulate its activity in a mutually exclusive manner, i.e. p47 in homotypic fusion and Ufd1/Npl4 in proteasomal function (24,25). It has been proposed that additional cofactors will be associated with other specific activities accounting for the ability of a single ATPase to mediate multiple functions (25). Presumably, these functions require precisely co-evolved protein⅐protein interactions among the ATPase and its various cofactors and substrates. It is surprising then that the TbVCP gene is able to complement a conditionally lethal yeast cdc48 mutant. Whether TbVCP can act as a global surrogate in yeast is not clear; we can only be certain that we have complemented the essential Cdc48p function(s) that are compromised in this particular mutant. Nevertheless, trypanosomes and fungi are widely divergent (2) and, despite 64% identity between the two homologues (1), it is remarkable for complementation to occur when multiple protein⅐protein interactions are required.
There is an emerging concept of how VCP homologues may function as force-generating motors. This view is perhaps best illustrated by the role of the mammalian homologue p97 and its cofactor p47 in homotypic membrane fusion. Fusion is thought to be driven by the interaction of SNARE proteins in opposing membranes after which p97⅐p47 complex mediates the ATP-dependent disruption of paired SNAREs in the post-fusion membrane. The complex would then dissociate leaving the free SNAREs available for further rounds of fusion. Indeed, p97⅐p47 has been shown to interact with the t-SNARE syntaxin 5 in both endoplasmic reticulum and Golgi membranes (22,23). Furthermore, binding is mediated by p47 (23) and after in vitro fusion, p97⅐p47 is released from membranes (22). Recent crystallographic and cryoelectronmicroscopy studies suggest a detailed structure for the p97⅐p47 complex (47,48). p97 monomers assemble as two stacked concentric rings composed of hexamers of the D1 and D2 sub-domains with six p47 subunits associated around the periphery. Binding of either ATP or ADP results in a conformational change in the core p97 structure. In the absence of nucleotide, or with bound ADP, p47 association is loose and mobile. But, strikingly, in the presence of ATP the p47 monomers project radially in a constrained manner. Overall, these findings suggest a model in which ATP binding induces a strained conformation in p97 that pushes p47 monomers outward from the central axis, which in turn could generate the force required to disassemble a t-SNARE pair attached to any two p47 monomers within the complex (48). Subsequent ATP hydrolysis would restore the relaxed conformation followed by release of p97⅐p47 from membranes.
Our findings with TbVCP mutants have significant implications for this model. The D2 hydrolysis motif is absolutely essential for in vivo TbVCP function and, although not essential, the D1 hydrolysis motif apparently modulates overall activity. Thus, ATP hydrolysis by the D2 domain is likely to be central to the observed conformational changes in the p97⅐p47 complex. Zhang et al. (47) have speculated that ATP hydrolysis in one domain promotes nucleotide exchange in the other, and that subsequent hydrolysis by the second domain promotes nucleotide exchange in the first, producing a ratchet-like conformational cycle for force generation in the p97 core. Our results argue against an essential role for ATP hydrolysis by the D1 domain in such a model, but a role for this domain in overall function cannot be discounted. Because it is bound ATP that induces the strained conformation in the p97⅐p47 complex (48), reversible nucleotide binding in the D1 domain may contribute to, or even be essential for, the strained-relaxed oscillation. Perhaps the affinity of the D1 binding site is low enough to allow reversible ATP binding without hydrolysis, or alternatively, the full hydrolysis/exchange cycle in the neighboring D2 domain may regulate ATP exchange in the D1 domain as suggested by Zhang et al. (47). In either scenario, ATP hydrolysis in the D1 domain, although not required, would enhance normal nucleotide exchange thereby accounting for the observed phenotypic differences between the DB2 and DB1/2 TbVCP mutants.
Obviously, proof and/or refinement of these models will require further work. Important issues that remain are whether both ATP binding sites must be occupied for conformational change, at what point nucleotide exchange occurs, and whether bound nucleotide is required for association of p97⅐p47 with SNARE complexes. Addressing these issues will require more structural studies utilizing mutant proteins such as we have generated. In addition, the establishment of trans-dominant mutants in an inducible trypanosomal expression system provides unique opportunities for further research into TbVCP function. In particular, with the loss of ATPase activity it may be possible to lock TbVCP in a bound conformation leading to identification of other essential protein cofactors and substrates in trypanosomes. Given the high sequence conservation of AAA proteins and the ability to complement genetically over wide phylogenetic distances, it is likely that any new findings will extend to other VCP homologues.