A Developmentally Regulated Aconitase Related to Iron-regulatory Protein-1 Is Localized in the Cytoplasm and in the Mitochondrion of Trypanosoma brucei *

Mitochondrial energy metabolism and Krebs cycle activities are developmentally regulated in the life cycle of the protozoan parasite Trypanosoma brucei. Here we report cloning of a T. brucei aconitase gene that is closely related to mammalian iron-regulatory protein 1 (IRP-1) and plant aconitases. Kinetic analysis of purified recombinant TbACO expressed inEscherichia coli resulted in a K m (isocitrate) of 3 ± 0.4 mm, similar to aconitases of other organisms. This was unexpected since an arginine conserved in the aconitase protein family and crucial for substrate positioning in the catalytic center and for activity of pig mitochondrial aconitase (Zheng, L., Kennedy, M. C., Beinert, H., and Zalkin, H. (1992)J. Biol. Chem. 267, 7895–7903) is substituted by leucine in the TbACO sequence. Expression of the 98-kDa TbACO was shown to be lowest in the slender bloodstream stage of the parasite, 8-fold elevated in the stumpy stage, and increased a further 4-fold in the procyclic stage. The differential expression of TbACO protein contrasted with only minor changes in TbACO mRNA, indicating translational or post-translational mechanisms of regulation. Whereas animal cells express two distinct compartmentalized aconitases, mitochondrial aconitase and cytoplasmic aconitase/IRP-1, TbACO accounts for total aconitase activity in trypanosomes. By cell fractionation and immunofluorescence microscopy, we show that native as well as a transfected epitope-tagged TbACO localizes in both the mitochondrion (30%) and in the cytoplasm (70%). Together with phylogenetic reconstructions of the aconitase family, this suggests that animal IRPs have evolved from a multicompartmentalized ancestral aconitase. The possible functions of a cytoplasmic aconitase in trypanosomes are discussed.

Trypanosoma brucei is a protozoan parasite in the blood and tissue fluids of mammals and causes two major tropical dis-eases, sleeping sickness in man and nagana in cattle. The flagellate is transmitted in subsaharan Africa between humans, livestock, and a huge reservoir of game animals by the tsetse fly, a bloodsucking insect vector. During cyclical transmission between mammals and tsetse, trypanosomes differentiate into a series of life cycle stages with distinct morphology, metabolism, and surface proteins to cope with the changing host environments and host defense mechanisms (1). The proliferating forms in the host blood have elongated and slender appearance (long slender forms) and rely entirely on glycolysis for energy production (2)(3)(4). Inefficient substrate utilization seems affordable at that developmental stage, due to glucose homeostasis in host blood. Hence, slender forms lack significant amounts of Krebs cycle enzymes (5), and their residual mitochondrion does not contribute to ATP production. At the peak of a parasitemic wave, slender forms differentiate into a quiescent, cell cycle-arrested stage with stumpy morphology (stumpy form) which has a more developed mitochondrion and expresses citric acid cycle activities and an incomplete electron transport chain (5)(6)(7)(8). In culture, differentiation to the stumpy stage is induced by a cell density sensing mechanism acting via the cAMP second messenger pathway (9,10). A population of stumpy forms, in turn, can be triggered to differentiate rapidly and synchronously to the next stage in the life cycle, the procyclic forms, which populate the fly midgut and rely on respiration of proline as their major energy source. Therefore, induction of Krebs cycle enzymes and of a respiratory chain in stumpy forms is regarded as a preadaptation to the fly environment (8). The signal for differentiation to procyclic forms in the midgut of the tsetse is unknown; however, the process can be triggered in culture by a temperature shift to 27°C or most efficiently by addition of cis-aconitate or citrate to the culture medium (11,12). The differentiation inducing effect of cis-aconitate and citrate is highly specific. The peculiar role of these substrates in triggering a developmental switch and the coordinate induction of the citric acid cycle enzymes during differentiation to the stumpy stage prompted us to identify and characterize the aconitase gene of T. brucei.
Aconitase (citrate(isocitrate) hydro-lyase, EC 4.2.1.3) catalyzes the stereospecific dehydration/rehydration reaction of citrate to isocitrate via the intermediate cis-aconitate (13). Activity of the enzyme critically depends on the presence of an iron sulfur [4Fe4S] cluster in the catalytic center that is highly sensitive to oxygen. Upon oxidation the cluster is converted to an inactive [3Fe4S] form (14). Mammalian cells express two distinct aconitases encoded by separate genes: (a) mitochondrial aconitase and (b) a bifunctional cytoplasmic aconitase that is identical to the iron-regulatory protein (IRP-1). 1 IRP-1 is an RNA-binding protein interacting with iron-responsive elements (IREs) in the untranslated regions of several mRNAs (15), including the transferrin receptor and ferritin H-and L-subunits (16), and functions in coordinate post-transcriptional regulation of cellular iron metabolism (17,18). In ironloaded cells, IRP-1 assembles a cubane [4Fe-4S] cluster, which renders it active as a cytosolic aconitase but inactive for RNA binding. Disassembly of the cluster upon iron starvation yields the active RNA-binding form that regulates stability or translation of target RNAs. The two conformations and functional states are mutually exclusive. Cytoplasmic aconitase activity might also be required for a glyoxalate cycle in specialized animal tissues like brown adipose tissue (19). In plants, where cytoplasmic aconitase is developmentally regulated and abundant in germinating seeds, its role in the glyoxalate cycle is well established (20,21).
Here we report on cloning and characterization of a protozoan aconitase that is closely related to animal IRPs and plant aconitases. Dual subcellular localization of the developmentally regulated T. brucei enzyme suggests that it not only takes part in the mitochondrial Krebs cycle but may have a yet unknown function in the cytoplasm of the parasite.
DNA Sequencing-The sequence of the antisense strand was determined from overlapping deletion clones of plasmid Tbaco1.1, constructed as recommended by the supplier of the Nested Deletion Kit (Amersham Pharmacia Biotech). The nucleotide sequences of the sense strand, of cDNA clones, and of the genomic clone obtained by inverted PCR were determined by primer walking. The Sequenase TM version 2.0 Sequencing Kit (U. S. Biochemical Corp.) and a fluorescent dideoxynucleotide terminator cycle sequencing kit (Applied Biosystems) were used for manual and automated sequencing, respectively.
Plasmid Constructs and Site-directed Mutagenesis-The complete coding region of TbACO was assembled from a 104-bp PstI-EcoRV fragment derived from a 3Ј-RACE clone and the genomic clone Tbaco1.1 opened with PstI (partial digest) and XbaI (blunted). For convenient cloning, a TbACO cassette was constructed as follows: a BamHI site was introduced immediately upstream of the start codon by recombinant PCR using the primers 5Ј CGGGATCCATGCTCAGCACGAT-GAAG 3Ј and JS5-2 (see cDNA cloning), and a BglII site was introduced exactly after the stop codon with the primers RIRP-2 (see cDNA cloning) and 5Ј GAAGATCTCTACAAATTACCCTTGATT 3Ј. The cassette was then recloned between BamHI and SalI of BluescriptSKII(Ϫ). Details of these constructions are available upon request.
For expression of C-terminally tagged TbACO in trypanosomes, the insertion vector pLew20 which targets the rDNA spacer of T. brucei (28) was opened with HindIII and BamHI; the ends were blunted, and the vector fragment was gel-purified. The Ty-1 epitope (29) was attached to the C terminus of TbACO by recombinant PCR using the TbACOspecific primer 5Ј ATGGCAGGAATCCACATA 3Ј and the Ty-1 primer 5Ј GCTCTAGACTAGTCAAGTGGATCCTGGTTAGTATGGACCTCCAAA-TTACCCTTGATTTTC 3Ј including an XbaI site. The resulting PCR fragment was cleaved with XbaI, blunted, and then cleaved with Eco47III. The TbACO cassette was cut with BamHI, blunted, and then cut with Eco47III, followed by isolation of the fragment. The three fragments were ligated together to give pLewacotag.
Southern Blot Analysis-Genomic DNA was prepared by phenol extraction, dialysis, and isopropyl alcohol precipitation. After electrophoresis in TAE buffer, agarose gels were blotted onto a GeneScreen TM nylon membrane (NEN Life Science Products) by capillary transfer. The hybridization and washing temperatures were adjusted to 5°C below the calculated melting temperature of specific DNA-DNA hybrids for high stringency conditions and to 40°C below the calculated melting temperature for low stringency conditions, using the buffers described in Boshart et al. (30). A TbACO-specific probe of defined length (nucleotides 1778 -2088 of Tbaco1.1) was generated by primer extension (31) with polylinker-specific primers (SK and KS) using as template an EagI-and ApaI-cut pBluescript plasmid containing the 311-bp PCR fragment.
Northern Blot Analysis-Total RNA was isolated using the guanidinium isothiocyanate/CsCl 2 cushion method (32) and was fractionated on 1.2% (w/v) formaldehyde-agarose gels and blotted onto Gene-Screen TM (NEN Life Science Products) nylon membrane. A riboprobe corresponding to nucleotides 1778 -2088 of Tbaco1.1 was prepared and hybridized as described (30)⅐ The blot was washed three times for 30 min in 0.1ϫ SSC, 1% (w/v) SDS at 65°C. For control, the same blot was probed with the random prime-labeled plasmid pR4 containing part of the rRNA locus (33)⅐ Bacterial Expression of TbACO-E. coli M15rep4 (34) transformed with pQE30/5 was grown in LB medium containing 100 g/ml ampicillin and 25 g/ml kanamycin up to an A 600 of 0.7, and TbACO expression was induced by addition of 1 mM IPTG for 3 h at 25°C. Bacteria were freeze/thaw lysed and sonicated in 50 mM Na x H y PO 4 buffer, 300 mM NaCl, pH 8.0, 40 g/ml bestatin, 1 g/ml leupeptin, 0.5 g/ml Pefabloc® (Roche Molecular Biochemicals) and 0.7 g/ml pepstatin. The soluble material was purified by metal chelate affinity chromatography on a Ni 2ϩ -NTA column (Qiagen) according to the manufacturer's protocols. The yield was in the range of 230 g of soluble protein from 1 liter of culture. For further purification, this protein was dialyzed against 20 mM Tris-HCl, pH 7.5, 5% (v/v) glycerol and subjected to ion exchange chromatography on a MonoQ HR 5/5 column (Amersham Pharmacia Biotech). The N terminus of the recombinant protein was verified by microsequencing; an amino acid analysis gave results compatible with the predicted sequence of recombinant TbACO, and a mass spectrum gave a value of 99902.5 mass units compared with a theoretical value of 99702.7 (all performed by the protein chemistry core facility of the MPI for Biochemistry).
Enzyme Assays-For aconitase activity measurements, recombinant full-length TbACO (99 kDa) was expressed and purified as described above, except that all buffers contained 10 mM citrate and 2 mM ␤-mercaptoethanol and were degassed and saturated with nitrogen. The aconitase substrate citrate protects from oxidative inactivation (35). For concentration and buffer exchange (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 10 mM citrate, 1 mM DTT), Centricon TM 100 ultrafiltration units (Amicon) were used. Protein samples were stored in a glove bag under argon or nitrogen atmosphere at 4°C. Iron loading under reducing conditions was performed in a glove bag with 200 g of protein (in 20 mM Tris-HCl, pH 7.5) by adding 10ϫ loading buffer (10 mM MES, pH 5.5, 10 mM ferrous ethylenediammonium sulfate, 50 mM DTT) and incubating for 25 min at 37°C similar as described by Kennedy et al. (36). The loaded protein was aliquoted and stored on ice in nitrogen atmosphere until measurement of activity with a double beam photometer (Perkin-Elmer Lambda 5). For the UV assay described by Henson and Cleland (37), 1 ml of 90 mM Tris-HCl, pH 7.5, containing DLisocitrate (Sigma) ranging from 1 to 20 mM was equilibrated at 25°C, and the formation of cis-aconitate was recorded for 6 min (extinction coefficient for cis-aconitate ⑀ ϭ 3.4 mM Ϫ1 cm Ϫ1 at 240 nm). Assuming Michaelis-Menten kinetics, nonlinear curve fitting of kinetic data was performed with Kaleidagraph 3.0.8 (Synergy Software) and the K m and V max values were calculated from a reciprocal plot according to Hanes (38). Aconitase activity in whole cell extracts was measured as described by Overath et al. (7). Protein concentrations were determined using the Bradford assay as supplied by Bio-Rad and bovine serum albumin as standard.
Antibodies-For the first series of immunizations (including "rat 1"), an 88-kDa recombinant TbACO fragment (pQE31/13) was purified by Ni 2ϩ -NTA chromatography, electrophoresed on an 8% (w/v) SDS-polyacrylamide gel, and blotted onto nitrocellulose membrane. The 88-kDa band was excised from the membrane (150 -200 g of protein), and the dried nitrocellulose was powdered, emulsified with complete Freund adjuvant, and injected intraperitoneally in female Wistar rats. Rats were boostered once with 150 g of protein after 4 weeks. Two weeks later, the animals were bled, and sera were affinity purified as described (39). For a second series of immunizations (including "rat 2" and "rabbit 1") the recombinant full-length TbACO (pQE30/5) was purified by Ni 2ϩ -NTA chromatography in soluble form and directly emulsified with complete Freund adjuvant. Rabbits were immunized by subdermal injection of 500 g of protein followed by 3 booster injections at regular intervals. Rabbit anti-phosphoglycerate kinase C antibodies were a kind gift of P. Michels. Rabbit anti-Synechococcus sp. HSP60 antiserum was purchased from StressGen Biotechnologies Corp. The BB2 hybridoma (anti-Ty1 monoclonal antibody) was a kind gift of P. Bastin and K. Gull. Rabbit anti-Dictyostelium ␣-tubulin antiserum was obtained from G. Gerisch.
Immunoblot Analysis-Whole cell extracts were prepared by transferring small volumes of PBS-washed and resuspended trypanosomes into SDS-lysis buffer preheated to 100°C. Samples were boiled for 5 min and sonified. Proteins were separated on 8% (w/v) SDS-polyacrylamide gels and were transferred onto Hybond TM polyvinylidene difluoride membranes (Amersham Pharmacia Biotech) using a semi-dry blot- The cDNA at the 3Ј-end (reverse transcriptase-PCR, GenBank TM accession number AF127457) was amplified with nested gene-specific primers RIRP2 (3) and RIRP1 (4). Four independent cDNA clones contributed 75 amino acids of coding region (boxed) as well as 254 bp of 3Ј-UTR. The trans-splice site (marked by an AG dinucleotide) at position Ϫ100 relative to the first ATG was determined from the sequence of 3 independent 5Ј-RACE clones amplified with nested gene-specific primers JS5-1 (1) and JS5-2 (2). The 3Ј-end of the gene was cloned by inverted PCR (INV-PCR, GenBank TM accession number AF190556) using the nested primer pairs JS/5.4 (5) plus IRP.4 (6) and JS5-2 (2) plus IRP.1 (7). The 3Ј-ends of TbACO transcripts (wavy lines) were mapped by 3Ј-RACE PCR using the gene-specific primer IRP.6 (8). Two PCR products of 500 and 810 bp (indicated by arrows) were obtained, thus calculated sizes of TbACO mRNAs (without poly(A)-tail) of 4197 and 4507 nt, respectively. The transcript sizes estimated from Northern blots (see Fig. 6) are indicated (4.2 and 4.7 kb).
Digitonin Fractionation-Cell fractionation was essentially done as described by Hä usler et al. (41). Aliquots of 2.5 ϫ 10 7 procyclic cells were washed in 25 mM Tris-Cl, pH 7.8, 1 mM EDTA, 0.6 M sucrose, 1 mM DTT, 2 g/ml leupeptin and were resuspended in 1.12 ml of the same buffer, and 125 l of prediluted digitonin was added while mixing. After 2 min at 37°C and 10 s on a Vortex mixer the cells were centrifuged at 12,000 ϫ g at 4°C for 10 min. The supernatant was concentrated by ultrafiltration using Microcon TM 30 (Amicon) before SDS-PAGE and immunoblotting of aliquots of both fractions.
Immunofluorescence-Short stumpy bloodstream forms were freshly isolated from infected mouse blood by centrifugation (22) and were stained with the mitochondrion-selective dye MitoTracker green FM TM (Molecular Probes) as described (42). Cells were then fixed in 3% (w/v) paraformaldehyde in PBS, permeabilized with 0.1% (v/v) Triton X-100 in PBS, and stained for 1 h with rabbit anti-TbACO antiserum or preimmune serum (Fig. 7A) diluted 1:10 in PBS, 1% (w/v) bovine serum albumin. Texas Red TM -conjugated goat anti-rabbit IgG F(abЈ) 2 fragments (Dianova) were used as secondary antibodies. Washed cells were mounted in VectaShield TM (Vector) on 3-aminopropyltriethoxy-silanecoated slides and were examined with a Zeiss Axiophot 2 microscope equipped with a Plan-Apochromat 63ϫ objective (NA 1.4). Images were recorded with a cooled digital CCD camera (MicroMAX, Princeton Instruments).
Evolutionary Analysis-The multiple sequence alignments were gen-erated with ClustalW, version 1.60 (43), and manual editing. The total length of the alignment comprises 1171 sites and is available upon request. To reconstruct phylogenetic trees based on the amino acid sequence alignment, the PUZZLE program (version 4.0) was applied (44,45). As substitution model the substitution matrix BLOSUM 62 (46) was assumed. To account for rate heterogeneity a discrete Gamma model (47) with four categories was introduced, and the corresponding shape parameter was estimated. For each sequence we tested if the base composition deviates significantly from the average composition. The estimated shape parameter of the gamma distribution equals 1.17, thus indicating weak rate heterogeneity.

RESULTS
Cloning of an Aconitase Gene of T. brucei-Highly degenerate pools of PCR primers matching two conserved sequence motifs of the aconitase gene family were used to amplify genomic T. brucei DNA. Ten out of 31 clones derived from a PCR product of expected size contained the same 311-bp insert with an uninterrupted open reading frame exhibiting 74% protein sequence similarity to human IRP-1 and 75% similarity to Arabidopsis thaliana aconitase. No other aconitase family-related sequence was detected with this and other pairs of degenerate primers. The cloned 311-bp fragment was used to probe Southern blots of genomic DNA. Each restriction enzyme digest produced a single specific band with different mobility, indicating a single copy gene (Fig. 1). Low stringency hybridization conditions (40°C below T m ) revealed only one additional sequence (marked with asterisks in Fig. 1) which was cloned from a phage library. The cross-hybridizing fragment did not exhibit any nucleotide sequence similarity to the aconitase gene family. Thus, the hybridization conditions used must have been sufficiently relaxed and sensitive for detection of TbACO-related genes. A restriction map of the TbACO locus was constructed from multiple digests, and a genomic 3.6-kb  (110) entry 7ACN was adopted for pig heart aconitase. The alignment was compiled by ClustalW, version 1.60 (43), with default settings (gap opening penalty of 10 and gap extension penalty of 0.05). The presequence of porcine mitochondrial aconitase was not included, and manual adjustments of the alignment were made at the N terminus. Sequence identity with TbACO is indicated by gray background, and active site residues (see Table I) are highlighted in black with white lettering. Additional residues highly conserved in the FeS isomerase family (53) are marked by an asterisk. The motifs initially chosen for design of degenerate PCR primers (see "Experimental Procedures") are underlined.
SacII-PstI fragment was cloned (plasmid Tbaco1.1, see "Experimental Procedures" and Fig. 2) which contained a long open reading frame devoid of a stop codon. Therefore, a cDNA containing the missing C terminus was cloned by reverse transcriptase-PCR with nested TbACO-specific primers. An 800-bp internally primed cDNA was obtained. Therefore, the 3Ј-end of the gene was cloned by inverted PCR (see Fig. 2 and "Experimental Procedures"). The length of the 5Ј-UTR of the TbACO transcript was determined by a modified 5Ј-RACE strategy exploiting the conserved mini-exon sequence present at the 5Ј-end of all trypanosomal mRNAs. The sequence of three independent clones derived from the single PCR product showed that the mini-exon was added at a trans-splice site (AG dinucleotide) at position Ϫ100 relative to the first ATG codon. As expected for a canonical trans-splice site, several polypyrimidine stretches were found within 50 bp upstream of the AG.
The TbACO Protein Sequence-The open reading frame predicted a protein of 897 amino acids with a calculated molecular mass of 98,302 Da and a pI of 6.48. Both methionines at positions 1 and 5 are in sequence environments compatible with the requirements for translation initiation (48). Thus, use of the first ATG is assumed. Pairwise alignments of TbACO with other members of the aconitase family show a very high degree of amino acid identity and similarity with vertebrate IRPs and with plant aconitases (58/74% with human IRP-1; 61/75% with A. thaliana aconitase) extending over the entire length of the proteins. In contrast, mitochondrial aconitases of mammals or S. cerevisiae are significantly more distant (30% identity, 55% similarity). A multiple sequence alignment of TbACO with a representative member of each aconitase subfamily is shown in Fig. 3. From the crystal structure and mutation analysis of porcine mitochondrial aconitase, 24 residues were identified as important for coordination of the [4Fe-4S] cluster, for substrate recognition and catalysis, and for support of active site chains by hydrogen bonds (49 -52). Table  I (53). In TbACO these positions are substituted by methionine (Met-222) and phenylalanine (Phe-100), respectively, like in mammalian IRPs and plant aconitases (see Fig. 3). Arginine 580 that appears to be a key residue for substrate binding in pig mitochondrial aconitase (52) has been replaced by leucine (Leu-702) in TbACO. This position is only substituted in two other aconitase sequences, that of Legionella pneumophila (54) and a hypothetical aconitase of S. cerevisiae (hACO) identified by genome sequencing (55). A multiple alignment of Frishman and Hentze (53) has identified 26 additional residues that are highly conserved in the Fe-S isomerase family (marked with an asterisk in Fig. 3). Of these, only one glycine (Gly-413 of pig mitochondrial aconitase) is substituted by arginine (Arg-504) in the TbACO sequence.
The high degree of sequence similarity of TbACO with IRPs, and the fact that trypanosomes represent the earliest branch of extant eukaryotes that harbor mitochondria (56), solicited a phylogenetic analysis. Sequences of 22 aconitases and IRPs available in data bases were aligned. As outgroup, sequences served 10 isopropylmalate isomerase and bacterial aconitase B group sequences that are significantly different. A PUZZLE tree based on 32 aligned sequences was reconstructed. For the 22 aconitase and IRP sequences a PUZZLE tree was computed and rooted according to the big tree (Fig. 4). The general topology of the tree is very similar to published trees of the aconitase family (53,57,58). TbACO is clearly separated from the mitochondrial aconitases. Within the IRP-like subfamily, TbACO, plant aconitases, bacterial aconitases, and IRPs all branch from one node, and it is not possible to resolve the branching pattern any further. The statistical support for any particular topology linking TbACO to either the IRP cluster or the plant aconitase cluster is weak. This is congruent with recent evidence for an overestimation of the divergence of protists and crown eukaryotes, based on nuclear gene data, and the proposal of a "big bang" radiation of the various eukaryotic lineages (reviewed in Ref. 59).

Enzymatic Activity and Kinetic Properties of Recombinant TbACO-
The complete open reading frame of TbACO was fused to a hexahistidine tag at the N terminus, and the fusion protein was expressed in E. coli and purified by affinity chromatography on a metal chelate resin (Fig. 5A). Aconitase activity increased linearly with the amount of recombinant protein and was maximal between pH 7 and 8 (not shown). The activity was dependent on loading of the protein with Fe 2ϩ in the presence of DTT and was sensitive to oxygen (Table II), consistent with the presence of a labile [4Fe-4S] cluster (14,36,60). The conserved cysteine positions of TbACO ligating this putative iron-sulfur cluster (see Table I) were then substituted by serines by site-directed mutagenesis. Two recombinant mutant proteins (C446S and C512S/C515S) were purified and iron-loaded exactly as the wild type protein. As expected, no enzymatic activity could be detected (Table II and data not shown). The analogous mutations in human IRP-1 also abolished enzymatic activity (61,62). This control provided unambiguous evidence that we had cloned an aconitase. The specific activity under V max conditions was determined using Ͼ95% pure recombinant enzyme prepared by further purification on a MonoQ TM ion exchange column (Fig. 5A). The result was close to specific activities reported for bovine heart mitochondrial  (64)). The Michaelis-Menten constant (K m ) for the substrate isocitrate was determined from reaction velocity plots and derived reciprocal Hanes plots as shown in Fig.  5B. The resulting K m of 3 Ϯ 0.4 mM is about 1 order of magnitude higher than published values for pig heart mitochondrial aconitase (65). This can be correlated with the substitution of an active site residue implicated in substrate binding (Leu-702 in TbACO substituting Arg-580 in pig mitochondrial aconitase, see Table I). In fact, an R580K mutant of pig mitochondrial aconitase (52) exhibits a 30-fold reduced K m , nearly complete loss of tight substrate binding, and a 250 -30,000-fold decreased activity, depending on the substrate. The ratio of the activities with isocitrate and citrate is 2.1 for the wild type but 11 for the R580K mutant (52). For TbACO, this ratio was 1.6 and the specific activity was close to that of wild type pig mitochondrial aconitase (Table II). This is even more surprising as the positive charge of the residue, which is thought to interact with the ␥-carboxyl of substrates, is maintained in the R580K mutant of pig mitochondrial aconitase but replaced by the hydrophobic side chain of a leucine in TbACO. Thus, it seems that in TbACO a slightly different network of hydrogen bonds holds the substrate in place and substitutes for the proposed essential function of Arg-580 in the prototype aconitase structure (51,66).
Differential Expression in the Life Cycle-A major TbACO transcript of 4.7 kb and a minor transcript of 4.2 kb were detected by Northern hybridization in three life cycle stages of T. brucei. The use of two polyadenylation sites was confirmed by mapping the cDNA with 3Ј-RACE-PCR (Fig. 2). No difference in total mRNA abundance was noted between stumpy bloodstream forms and procyclic forms, and the sum of the two transcripts was only slightly reduced in slender bloodstream forms (Fig. 6). Regulation appeared to be more prominent for the smaller mRNA; however, its contribution to the total amount of aconitase-specific mRNA was less than 30%. To investigate TbACO protein expression, antisera were produced in rats and rabbits by immunization with purified recombinant TbACO. Irrespective of the animal species and immunization procedure (see "Experimental Procedures"), one single polypeptide of 98 kDa was detected in procyclic whole cell lysates, in agreement with a predicted molecular mass of 98,302 Da. No signals were detected by the respective nonimmune or preimmune sera (Fig. 7A). TbACO expression was then compared in slender and stumpy bloodstream populations and in procyclic forms of the pleomorphic strain AnTat 1.1. The maximal amount of protein was detected in the procyclic stage (Fig. 7B). Expression in the slender bloodstream stage was less than 4% of the procyclic level, and stumpy populations were found to be intermediate. The same result was obtained with trypanosomes harvested from several independent rodent infections and upon probing with different antisera (not shown). The absolute TbACO abundance was estimated by comparison with standard dilutions of purified recombinant TbACO in the Western blot shown in Fig. 7C and calculated to be close to 160,000 molecules per procyclic cell. The fraction of stumpy forms in the predominantly slender or predominantly stumpy populations was determined by cytochemical staining for the stumpy marker activity NADH dehydrogenase (6) (see legend to Fig. 7). After correction, an estimation of 40,000 molecules per stumpy cell and 5000 molecules per slender cell was reached. In summary, developmental regulation of TbACO is more than 30fold. This contrasts with a very minor change in mRNA expression. Thus, developmental control of expression must operate at the translational or post-translational level.
Subcellular Localization-The high sequence similarity of TbACO to IRPs, which are cytoplasmic regulatory proteins in mammalian cells, raised the question of its subcellular localization. TbACO was present but not enriched in a partially purified mitochondrial vesicle fraction prepared according to Ref. 67 (a kind gift of H. U. Goeringer, data not shown). This suggested multicompartmentalization of the protein. Therefore, procyclic trypanosomes were fractionated after differential permeabilization with digitonin. Based on different choles-terol content of the membranes, only the plasma membrane is permeabilized at low digitonin concentrations, whereas organellar membranes require higher concentrations (68 -70). Phosphoglycerate kinase B and the mitochondrial heat shock protein HSP60 were used as markers for the cytosolic and mitochondrial fractions, respectively, as described before (41). Procyclic cells were incubated with increasing concentrations of digitonin, followed by centrifugal separation of soluble and particulate fractions. The amount of TbACO and of the marker proteins in each fraction was quantified by densitometric scanning of Western blots (Fig. 8, A and B). About 70% of total TbACO was released into the supernatant together with the cytosolic marker phosphoglycerate kinase B at 0.5 mg of digitonin per mg of cellular protein. The remaining part of TbACO appeared to be as resistant to solubilization as the mitochondrial matrix marker HSP60. To exclude the possibility that our antibodies cross-reacted with a different 98-kDa protein in one of the cellular compartments, TbACO was epitope-tagged at the C terminus with a short peptide derived from the yeast transposable element Ty1 (29). The tagged open reading frame was inserted into the trypanosomal expression vector pLew20 (28) and targeted to the ribosomal spacer region of the T. brucei genome by homologous recombination. The resulting procyclic cell line, which expressed about the same amount of tagged TbACO as wild type protein (not shown), was subjected to digitonin fractionation using the same conditions as for wild type procyclic forms. Western blots were probed with an epitope tag-specific monoclonal antibody (Fig. 8, C and D). About 70% of the tagged TbACO was released together with the cytosolic marker (phosphoglycerate kinase B) at low digitonin concentrations where the mitochondrial HSP60 was quantitatively retained in the pellet fraction. Together, these experiments proved that about 70% of TbACO was localized in the cytoplasm of procyclic T. brucei. Mitochondrial localization of the remaining 30% was suggested by the fractionation; however, the protease digestion control required to rule out nonspecific association of TbACO with the particulate fraction was technically difficult due to relative resistance of free TbACO to digestion. Instead, localization of TbACO to the single mitochondrion of trypanosomes was directly documented by immunofluorescence microscopy. In the stumpy stage, the mitochondrion has an easily discernible tubular structure extending over the entire length of the cell and can be visualized with the aldehyde-fixable membrane potential sensitive fluorescent dye MitoTracker TM Green FM (42). Upon immunofluorescent staining with three different TbACO-specific antisera, perfect colocalization of the antibody signal with the MitoTracker mitochondrial marker was observed in every cell ( Fig. 9 and not shown). Double immunofluorescent staining with TbACO antibodies and antibodies detecting the mitochondrial matrix protein HSP60 also showed colocalization and confirmed antibody FIG. 6. Developmental profile of TbACO mRNA expression. A Northern blot with 5 g of total RNA isolated from a long slender (LS) bloodstream form population, a short stumpy (SS) bloodstream form population, and cultured procyclic forms (PCF) of strain Antat1.1 was probed with a riboprobe derived from the coding region of TbACO. In the SS population, 90% of the cells expressed the stumpy cell marker NADH dehydrogenase, as assayed by cytochemical staining (6). Equal loading of samples was internally controlled by rehybridization of the blot with a rDNA probe detecting the 18 S rRNA and the ␣and ␤-segments of 28 S rRNA. The relative abundance of the sum of the two TbACO transcripts marked with arrowheads was determined by densitometry and normalization to the rRNA signals (relative values are 1, 1.57, and 1.77 for LS, SS, and PCF, respectively). The size marker is an RNA ladder from Life Technologies, Inc. access to the mitochondrial matrix under the given permeabilization conditions (data not shown). TbACO-specific cytoplasmic staining was weak, as expected from leakage of cytoplasmic protein during permeabilization necessary for detection of mitochondrial proteins (Fig. 9). In conclusion, TbACO has a dual subcellular localization in the cytosol and in the mitochondrion of T. brucei. Cytosolic and mitochondrial aconitase activity was then measured in lysates after preparative fractionation of procyclic trypanosomes with 0.5 mg of digitonin per mg of total protein. The fractionation was controlled with the marker proteins phosphoglycerate kinase B and HSP60, and the enzymatic assay showed that 78 Ϯ 9% of the activity was in the cytoplasmic fraction and 21 Ϯ 2% was associated with organelles. Thus, the subcellular distribution of aconitase activity was in agreement with the distribution of TbACO protein.

DISCUSSION
Mammalian cells have two aconitases encoded by separate nuclear genes: (a) the mitochondrial citric acid cycle enzyme and (b) a cytoplasmic aconitase, better known as the ironregulatory protein (IRP-1) which acts as an iron sensor and post-transcriptional regulator and as a signal transducer for oxidative stress (71,72). Since members of both aconitase subfamilies were identified in invertebrates, e.g. Caenorhabditis elegans and Drosophila melanogaster (73), specialization of mitochondrial and cytoplasmic aconitase functions seems to have occurred early during animal evolution. Here we report on a protozoan aconitase which belongs to the IRP-1 subfamily but localizes in the mitochondrion as well as in the cytoplasm. TbACO accounts for total aconitase activity in T. brucei. This was directly confirmed by enzyme assays in trypanosome lines with targeted disruption of both alleles of TbACO. In several independent ⌬aco::HYG/⌬aco::NEO procyclic lines, 2 no aconitase activity could be detected (data not shown). The sensitivity of the enzymatic assay does not exclude a minor activity; however, three independent methods, PCR amplification with degenerate primers, low stringency Southern hybridizations, and Western blotting with several TbACO antisera ( Fig. 1 and Fig.  7, and not shown), did not indicate a second aconitase or a second IRP-related gene in T. brucei.
The mitochondrial localization of about 30% of TbACO and the absence of a second aconitase indicate that TbACO functions in the citric acid cycle of T. brucei. To our knowledge, this is the first direct evidence that an IRP-like aconitase functions in mitochondrial metabolism in a eukaryote. The dual subcellular localization resembles the situation in plant tissues, where 50 -90% of total aconitase activity or protein reside in the cytosol and function in the glyoxalate cycle (20,21,74,75). Plant aconitase purified from cytosol and mitochondria is indistinguishable with respect to kinetic parameters, molecular mass (90)(91)(92)(93)(94)(95)(96)(97)(98), and the EPR spectrum (20, 64, 74 -76). Furthermore, the molecular mass and the EPR spectrum of plant aconitase purified from mitochondria resemble mammalian IRP but not mammalian mitochondrial aconitase (64,77). The three plant aconitase sequences available so far (21,78) are very similar to IRPs and to TbACO but less related to the mitochondrial aconitase subfamily. Hence, it seems likely that the mitochondrial aconitase in plants is encoded by these IRPrelated sequences, although direct evidence has yet to be provided. Particularly, it is unknown whether the cytosolic and mitochondrial plant aconitase isoforms, which are differentially regulated and chromatographically separable (20), derive from a single or distinct genes. The dual localization of the T. brucei aconitase, which is similar to plant aconitases (Fig. 3), suggests that cytosolic and mitochondrial IRP-related isoforms may indeed be encoded by one gene in plants.
As the first IRP-like sequence from a protist, the TbACO sequence has added valuable phylogenetic information with respect to evolution of the aconitase gene family (53,57,58,79). The fact that TbACO forms a well supported clade together 2 B. Fast and M. Boshart, unpublished data.  2, rabbit 1). The equivalent of 5 ϫ 10 6 procyclic trypanosomes per lane was run on 8% (w/v) SDS-PAGE and probed with nonimmune serum (ni, from a control animal of the same rat strain) or immune serum (i), or matched preimmune serum (pi) at the following dilutions: 1:500 for affinity purified sera from rat 1 and rat 2 and affinity purified controls; 1:100 for serum and control from rabbit 1. For detection, the ECL kit (1st to 4th lanes) or an alkaline phosphatase-catalyzed color reaction (5th and 6th lanes) were used. B, lysates (two independent preparations for each developmental stage) of 5 ϫ 10 6 long slender forms (LS, lane 1), 2 ϫ 10 6 long slender forms (LS, lane 2), 2 ϫ 10 6 short stumpy forms (SS, lanes 3 and 4), or 2 ϫ 10 6 procyclic forms (PCF, lanes 5 and 6) of AnTat1.1 were size-fractionated by SDS-PAGE. The developmental stage of bloodstream form populations was verified with an established stumpy marker, NADH dehydrogenase, using cytochemical staining according to Vickerman (6). The slender populations contained Ͻ3% positive cells, and the stumpy populations contained Ͼ85% positive. TbACO was detected with purified rabbit 1 antibody (1:1000), and tubulin (TUB, internal control) was detected with rabbit anti-(Dictyostelium ␣-tubulin) serum (1:1000), followed by 125 I-protein A. Relative normalized intensities of TbACO bands as derived from PhosphorImager scans of 125 I and the fold regulation (average of n ϭ 2) are indicated below the lanes. All samples were run on the same gel. C, absolute quantification of TbACO by Western blotting of a lysate from PCF together with dilutions of a highly purified rTbACO standard (see under "Experimental Procedures"). Affinity purified antibodies from rat 2 (1:500) and the ECL kit were used for detection. Relative intensity values derived from densitometry are indicated below the lanes. Linear regression analysis of the scanned standard signals was used for calculation. with animal IRPs, plant aconitases, and bacterial aconitases of the E. coli acoA type, clearly separated from mitochondrial aconitases of yeast and animals (Fig. 4), strengthens the argument for the existence of aconitase paralogues before the separation of eubacteria from eukaryotes. The mitochondrial localization of TbACO and the branching of IRP, TbACO, and bacterial aconitases from one node support the view that the ancestor of the regulatory protein IRP was an enzyme of mitochondrial energy metabolism which most likely was acquired from a proteobacterial endosymbiont. It seems that members of the IRP-like and of the classical mitochondrial aconitase subfamilies both can function in more than one subcellular compartment. In S. cerevisiae, which does not harbor an IRPrelated gene in its genome (80), the mitochondrial aconitase is also present in the cytosol where it participates in the glyoxalate cycle (81). In trypanosomes, the single IRP-like aconitase is present in the mitochondrion and in the cytosol, and in plants, IRP-related aconitases seem to function in mitochondrial metabolism as well as in the cytosolic step of the glyoxalate cycle (20,21,75). Animals may have kept or acquired members of both aconitase subfamilies during evolution and restricted them to one subcellular compartment in order to specialize them for different tasks in the mitochondrion and in the cytosol.
The distribution of one gene product between different subcellular compartments can be achieved by a variety of mechanisms, most of which lead to alternative transcription and/or translation initiation sites or alternative mRNA processing (reviewed in Ref. 82). In the case of TbACO, there is only one defined 5Ј-end of the mRNA and two closely spaced potential AUG codons at positions 1 and 5. It is difficult to imagine alternative translation initiation at these positions to include a mitochondrial import presequence in a fraction of the synthesized protein. Furthermore, the N-terminal sequence of TbACO does not form an amphipathic helix and does not predict a classical mitochondrial targeting signal. The mechanisms for protein import in trypanosomes seem to be largely conserved (41,(83)(84)(85)(86); however, a second class of much shorter, typically 9 amino acid long presequences that resemble hydrogenosomal import sequences has been identified (87). At least one imported protein (cytochrome c 1 ) lacks a presequence (88), and a nonconservative import pathway has been suggested (86). Mitochondrial and cytoplasmic TbACO were never resolved as double band upon PAGE of T. brucei lysates, and hence, there is no evidence for cleavage of a leader sequence. However, cleavage of both mitochondrial and cytosolic mature protein as shown for S. cerevisiae fumarase FUM1 (89) cannot be excluded. All fumarase molecules synthesized in yeast are processed by the mitochondrial matrix protease, but nevertheless most of the enzyme (80 -90%) ends up in the cytosol. An aborted translocation process has been suggested to be responsible for the dual mitochondrial and cytoplasmic localization of S. cerevisiae fumarase FUM1 (89). It should be noted that the 9 amino acid presequence of T. brucei dihydrolipoamide dehydrogenase which is sufficient for import (85,87) is cleaved off in both bloodstream and procyclic forms, although the protein is not present in the mitochondrion of slender bloodstream forms (90 -93). We suggest that multicompartmentalization of TbACO is due to inefficient targeting or an aborted translocation mechanism, possibly associated with a weak import signal FIG. 8. Subcellular localization (digitonin fractionation). Procyclic trypanosomes were subjected to differential permeabilization by 0 -10 mg of digitonin/mg of cellular protein (concentration indicated on top of the lanes in A and C, a different batch of digitonin was used for C) and were subsequently fractionated by centrifugation. Supernatant (S) and pellet (P) fractions were run on 8 or 10% (w/v) SDS-polyacrylamide gels and were analyzed by Western blotting using anti-phosphoglycerate kinase C serum (1:2000), anti-HSP60 serum (1:2000), affinity purified rat anti-TbACO serum (1:300), and anti-Ty1 monoclonal antibody (BB2 culture supernatant, 1:50) as indicated. Phosphoglycerate kinase B (PGK-B, detected by anti-phosphoglycerate kinase C serum) served as cytosolic marker and mitochondrial heat shock protein HSP60 served as mitochondrial marker (41). A and B, AnTat1.1 wild type procyclic trypanosomes (2.5 ϫ 10 6 or 5 ϫ 10 6 cells (for HSP60 only) per lane). C and D, transgenic MiTat1.4 procyclic trypanosomes stably expressing an epitope (Ty1)-tagged TbACO. Western blots are shown in A and C, and the corresponding quantitative data obtained by densitometric scanning are displayed in B and D. The released fraction (% release) is calculated by dividing the amount of specific protein in the supernatant by the total amount of specific protein in supernatant and pellet of the respective sample. deviating from the presequence consensus features.
Why do trypanosomes have a cytoplasmic IRP-like aconitase? First, it is possible that the protein serves a gene regulatory function similar to mammalian IRP. Recently, the Bacillus subtilis IRP-like aconitase has been shown to bind to mammalian IREs (94). The sequence motif DLVIDH-IQV implicated in RNA binding of IRP (95) is nearly conserved in TbACO (three conservative substitutions); however, no sequence-specific binding of the mammalian IRE consensus RNA sequence to recombinant purified TbACO could be detected. 2 The putative trypanosomal target RNA sequence may deviate, and thus a regulatory role can only be addressed genetically. We have investigated expression and regulation of one possible target, the trypanosomal transferrin receptor, and found no change of its expression or regulation in cells carrying a targeted deletion of TbACO (96). Second, we have considered the possibility that TbACO serves as cytoplasmic iron store. In fact, the IRP-related aconitase of L. pneumophila is the major ironcontaining protein in that organism (54). However, a calculation based on the iron uptake rate of trypanosomes (97) showed that TbACO should contribute less than 10% to the total iron content of proliferating slender bloodstream forms. Given the abundance of cytoplasmic TbACO, particularly in the procyclic stage, a metabolic function seems the most likely. In plant tissues, cytoplasmic aconitase is developmentally regulated with a dramatic increase during seed and pollen maturation and during germination, reflecting glyoxalate cycle activity (reviewed in Ref. 19). Evidence for a glyoxalate cycle in stationary promastigotes (probably metacyclic forms) of the trypanosomatid Leishmania has been reported (98,99). Although key enzymes of the glyoxalate cycle have not been detected in procyclic forms of T. brucei in culture, 3 a functional glyoxalate cycle may be operative and important at some later stage in the insect vector, where utilization of storage lipids may compensate for temporary shortage of nutrients. We are currently testing this hypothesis by tsetse fly passage of TbACO knock out strains.
Citric acid cycle activities and certain respiratory chain activities are turned on in the stumpy bloodstream stage of T. brucei as a preadaptation to the tsetse midgut environment, where rapid differentiation to the procyclic stage is essential for survival (5,8). The developmental profile of TbACO expression fully accounts for the previously reported changes of total aconitase activity (5, 7) and parallels the developmental changes in energy metabolism in the trypanosomal life cycle (1,2). More than 30-fold developmental regulation of TbACO contrasts with only a minor change of mRNA abundance, indicating a translational or post-translational mechanism. The only other citric acid cycle enzyme that has been cloned so far, malate dehydrogenase, seems to be regulated in a similar fashion (100). We anticipate that a common translational or posttranslational mechanism may coordinately up-regulate citric acid cycle activities upon differentiation. Whereas numerous examples suggest a predominance of regulation at the level of differential mRNA stability in trypanosomatids (101,102), only a handful of examples of translational or post-translational regulation have been reported (103)(104)(105)(106)(107).