Structural and Biochemical Analyses of the Eukaryotic Heat Shock Locus V (HslV) from Trypanosoma brucei*

Background: A eukaryotic HslV (TbHslV) protease and two potential HslU (TbHslU1 and TbHslU2) ATPases have been isolated from Trypanosoma brucei. Results: We determined the crystal structure of TbHslV at 2.4 Å resolution. Only TbHslU2 activated TbHslV protease activity. Conclusion: A key tyrosine residue in TbHslU2 required for activating TbHslV was identified. Significance: This study lays the groundwork for understanding the eukaryotic HslVU system. In many bacteria, heat shock locus V (HslV) functions as a protease, which is activated by heat shock locus U (HslU). The primary sequence and structure of HslV are well conserved with those of the β-subunit of the 20 S proteasome core particle in eukaryotes. To date, the HslVU complex has only been characterized in the prokaryotic system. Recently, however, the coexistence of a 20 S proteasome with HslV protease in the same living organism has been reported. In Trypanosoma brucei, a protozoan parasite that causes human sleeping sickness in Africa, HslV is localized in the mitochondria, where it has a novel function in regulating mitochondrial DNA replication. Although the prokaryotic HslVU system has been studied extensively, little is known regarding its eukaryotic counterpart. Here, we report the biochemical characteristics of an HslVU complex from T. brucei. In contrast to the prokaryotic system, T. brucei possesses two potential HslU molecules, and we found that only one of them activates HslV. A key activating residue, Tyr494, was identified in HslU2 by biochemical and mutational studies. Furthermore, to our knowledge, this study is the first to report the crystal structure of a eukaryotic HslV, determined at 2.4 Å resolution. Drawing on our comparison of the biochemical and structural data, we discuss herein the differences and similarities between eukaryotic and prokaryotic HslVs.

In prokaryotes, two-component ATP-dependent proteases, such as HslVU, 2 ClpAP, and ClpXP, act as protein quality controllers via destruction and recycling of misfolded or damaged proteins (1). In the ATP-dependent proteases, the ATPase unfolds and translocates substrates, whereas the protease degrades the unfolded substrate. The protein degradation mechanism of a large eukaryotic protease complex, termed the 26 S proteasome, is similar to that of prokaryotic ATP-dependent proteases (1)(2)(3).
The heat shock protein complex, HslVU, is a simple homolog of the eukaryotic proteasome (4,5). In many bacteria, heat shock locus V (HslV) functions as a protease with its activator heat shock locus U (HslU), which is an unfoldase driven by ATP hydrolysis. According to MEROPS classification (see the MEROPS Web site), both HslV and the proteasome contain an N-terminal threonine that acts as the essential catalytic residue (6). They share ϳ20% primary sequence similarity, and the structure of HslV is well conserved with that of the ␤-subunit of the eukaryotic 20 S proteasome core particle (7). The dodecameric HslV, which resembles a "double donut" shape, forms a functional HslVU complex with two hexameric HslU molecules to seal the substrate entrance pores at both ends (4). The extensive biochemical study performed on the HslVU complex has rendered it a suitable model system that can be used to understand the eukaryotic 26 S proteasome (4, 5, 8 -14). Until recently, the structures of the individual components HslU and HslV as well as the HslVU complex have been studied in bacteria and archaea (7,(15)(16)(17)(18)(19)(20)(21)(22).
Although the HslVU has been detected in many prokaryotic systems, the coexistence of a proteasome with HslVU in a wide range of eukaryotic systems has recently been suggested (23)(24)(25). Among them, the HslVU from the protozoan parasites, Leishmania, Trypanosoma, and Plasmodium, which cause significant human diseases, including African sleeping sickness, malaria, and leishmaniasis, have been experimentally characterized (26 -29). Intriguingly, HslU and HslV are localized in the mitochondria in Trypanosoma brucei, where they have a novel function in regulating mitochondrial DNA replication (28). The amino acid sequence of HslV from T. brucei (TbHslV) shares more than 40% identity with that of HslV from Escherichia coli (EcHslV; Fig. 1A). There are two potential HslU molecules, HslU1 and HslU2, in T. brucei (TbHslU1 and TbHslU2) that have high identity with the HslU from E. coli (EcHslU; Fig.  1B) (28).  In order to understand the molecular features of the eukaryotic HslVU system, we performed biochemical characterization of TbHslV and TbHslU by using a synthetic substrate, benzyloxycarbonyl-Gly-Gly-Leu-7-amido-4-methyl coumarin (Z-GGL-AMC) (4,30) and a natural substrate of EcHslU, the SulA protein (31,32). Although both TbHslU1 and TbHslU2 regulate mitochondrial DNA replication (28), we found that only TbHslU2 acts as an activator of TbHslV protease. We also determined the first structure of the eukaryotic TbHslV. The general structural features of eukaryotic HslV are well conserved with those of the prokaryotic HslV. However, a specific interaction between TbHslU2 and TbHslV was detected from the combined structural and biochemical data obtained on TbHslV. Thus, this study lays the groundwork for understanding the eukaryotic HslVU system.

EXPERIMENTAL PROCEDURES
Cloning-The template DNA was a kind gift from Prof. Wang (University of California, San Francisco). The DNA coding for mature TbHslV was amplified by polymerase chain reaction (PCR) with forward and reverse primers containing sites for the restriction enzymes NdeI and XhoI. The PCR product was cloned into the pET-22b(ϩ) vector, and the resultant plasmid had a C-terminal hexahistidine tag and the essential catalytic threonine residue at its N terminus (Thr 1 ; Fig. 1A). EcHslU and SulA were amplified by colony PCR using E. coli DH5␣. The PCR product of EcHslU was cloned into pET-12a vector by using the NdeI and BamHI restriction enzyme sites, and the resulting construct contained an N-terminal octahistidine tag. The SulA construct was ligated into the pMAL-p4X vector by using the BamHI and HindIII restriction enzyme sites. All mutants, including T1A TbHslV and EcHslU containing the C-terminal sequence of TbHslU2 (EcHslU TbHslU2 ), were generated using the QuikChange site-directed mutagenesis technique (Stratagene). The subsequently obtained DNA sequences were confirmed by DNA sequencing.
Protein Overexpression and Purification-TbHslV was transformed into BL21(DE3)RIL cells. The transformed cells were cultured in LB medium containing 50 g/ml ampicillin and 34 g/ml chloramphenicol at 37°C until an A 600 nm reading of 0.5 was obtained. Expression was induced by adding isopropyl ␤-Dthiogalactoside to a final concentration of 1 mM at 16°C for 24 h. Cells overexpressing TbHslV were harvested by centrifugation, and the pellet was resuspended in 50 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 10% (w/v) glycerol and then subsequently disrupted by ultrasonication. The cell lysate was centrifuged, and the supernatant was applied to a nickel-chelating Sepharose column (GE Healthcare). Further purification was carried out by successive anion exchange (Mono Q TM 10/100 GL, GE Healthcare) and size exclusion (Superose TM 6 10/300 GL, GE Healthcare) chromatography. Eluents from columns were analyzed by SDS-PAGE and confirmed by N-terminal sequencing. The final protein solution was concentrated to 10 mg/ml in storage buffer (20 mM Tris-HCl (pH 7.7), 300 mM NaCl, 1 mM EDTA, and 1 mM NaN 3 ). The expression and purification of EcHslV (7), EcHslU (15,22), and MBP-SulA (33) have been described previously.
Activity Assay-Peptide hydrolysis was assayed using the chromogenic peptide Z-GGL-AMC (Bachem) as a substrate of TbHslV and EcHslV (34). EcHslU and various octapeptides derived from the sequence of the C-terminal segment of EcHslU, TbHslU1, and TbHslU2 were used for HslV activation. The activity assay was conducted at 37°C using storage buffer containing 7.5% (v/v) dimethylformamide in a total volume of 200 l, and the release of AMC was monitored as a fluorescence increment at 440 nm (excited at 360 nm) by using a Spectra-Max M5 system (Molecular Devices, Inc.) with a 96-well plate (Corning). For protein substrate degradation, the MBP-SulA was used as described previously (20,31,35).
Crystallization and Data Collection-TbHslV was crystallized using the sitting drop or hanging drop vapor diffusion method. In all cases, crystallization was performed at 22°C. For the hanging drop vapor diffusion method, the crystallization drop comprised 200 nl of protein and an equal volume of reservoir solution containing 0.1 M acetate (pH 5.5), 2.0 M ammonium sulfate, and 2% (w/v) PEG 400. The crystallization set-up was done by using a Mosquito crystallization robot (TTP LabTech, Melbourn, UK). The crystal was obtained within a day. The Form-I crystal belongs to monoclinic space group P2 1 with unit cell parameters of a ϭ 100.9 Å, b ϭ 107.0 Å, c ϭ 132.8 Å, and ␤ ϭ 104.3°.
For the hanging drop vapor diffusion method, each crystallization drop was mixed with 1 l of protein and an equal volume of reservoir solution, which contains 0.1 M Tris-HCl (pH 8.5) and 3.5 M sodium formate. Thus, the crystal Form-II was obtained in the orthorhombic space group I222, with cell parameters of a ϭ 105.9 Å, b ϭ 111.5 Å, c ϭ 117.2 Å, and ␣ ϭ ␤ ϭ ␥ ϭ 90°. The cryosolutions were 0.1 M acetate (pH 5.5), 2.0 M ammonium sulfate, 2% (w/v) PEG 400, and 20% (w/v) glycerol for Form-I crystal and 0.1 M Tris-HCl (pH 8.5) and 4.5 M sodium formate for the Form-II crystal. Before the crystals were cryocooled in liquid nitrogen, they were washed in cryosolutions.
Diffraction data were collected at the BL44XU beamline of Spring-8 (Hyogo, Japan) and the NW12 beamline of the Photon Factory (Tsukuba, Japan) by using an ADSC quantum chargecoupled device detector. A total of 180 images were collected with 1°oscillation, and each image was exposed for 0.8 s. The diffraction data were processed and scaled using the HKL2000 software package (36), and the statistics for the data collection are described in Table 1.
Structure Determination and Refinement-Phases were obtained by molecular replacement with the program MOLREP (37) in the CCP4 program suite (38). A previously determined structure of EcHslV was used as a search model (15). The initial model was rebuilt and refined using standard protocols in COOT (39), PHENIX (40,41), and REFMAC (42) until the R-factor was converged. During the refinement, noncrystallographic symmetry restraints were applied. The refinement statistics for the TbHslV structures are described in Table 1.
Assessment of model geometry and assignment of secondary structural elements were achieved using the program MOLPROBITY (43). Figures depicting molecular structures were generated using CCP4MG (44).
Molecular Modeling-The complex model between TbHslV and the C-terminal segment of TbHslU2 was generated using an HiHslVU complex structure (HslVU complex from Haemophilus influenzae) as a template (16). The coordinates of TbHslV are well superposed with those of HslV from H. influenzae (HiHslV) in the HiHslVU complex except for the residues from Gly 48 to Ala 93 . As a result of this superposition, the structure of TbHslV from residue 48 to 93 was replaced with the structure of HiHslV from the same corresponding residues. The alteration of this structure was then realigned with the sequence of TbHslV. The initial model was energy-minimized with SPBDV (45) and the CNS package (46).

RESULTS
Overproduction and Purification of Mature TbHslV-It is known that one HslV and two HslUs exist in T. brucei (28), which have N-terminal signal peptides that direct them to the mitochondria. Excluding the signal sequence, TbHslV shares 41.4% sequence identity with EcHslV and 46.3% sequence iden-tity with HiHslV (Fig. 1A). In addition, TbHslU1 and TbHslU2 share 43.0 and 40.6% sequence identity with EcHslU and 43.2 and 41.2% sequence identity with HslU from H. influenzae (HiHslU), respectively (28). The mature form, in which the signal sequence (residues 1-19) is removed, has the catalytic threonine residue at the new N terminus; the adjacent residue is also threonine (Fig. 1A), which is a distinct feature of N-terminal threonine proteases, including HslV as well as the 20 S proteasome (3,47). Indeed, there are also tandem threonine residues at the 8th and 9th positions (Fig. 1A). Therefore, we generated three different constructs (1-end, 8-end, and 20-end); only the TbHslV construct without the putative N-terminal 19-residue signal sequence was well expressed in E. coli, whereas the remaining constructs were expressed at low levels and were insoluble. Similar behavior was observed with full-length HslV from Thermotoga maritima (TmHslV), which belongs to the Archaea domain (20). We successfully purified C-terminal His 6 -tagged TbHslV, whose sequence begins from the essential catalytic threonine residue. The first methionine was clearly processed. The N-terminal amino acid sequence of purified TbHslV was Thr-Thr-Ile-Ser-Leu, confirmed by the N-terminal sequencing analysis. The oligomeric state of TbHslV is dodecameric in solution, which was confirmed by gel filtration analysis (data not shown). Therefore, the mature form of TbHslV exhibits the basic biochemical characteristics of prokaryotic and archaeal HslVs.
Only the C-terminal Peptide of TbHslU2 Activates the Peptidase Activity of TbHslV-It is known that HslV possesses basically no peptidase and protease activity without its activator HslU (5,15). For the peptidase activity assay, we tried to overexpress TbHslU1 and TbHslU2 in E. coli, but these proteins could not be obtained in soluble form. As an alternative, we checked TbHslV activity in the presence of EcHslU as in the case of CodW, the HslV from Bacillus with EcHslU, which shows cross-species reactivity (48). However, we did not observe activation of TbHslV by EcHslU. According to previous reports (8,16,30), EDLSRFIL octapeptide, an amino acid sequence derived from the C-terminal 8 residues of EcHslU, can activate EcHslV, as assayed using Z-GGL-AMC as a substrate. Therefore, we synthesized two different octapeptides corresponding to the C-terminal octapeptide regions of TbHslU1 and TbHslU2, respectively (Fig. 1B). Unlike the C-terminal peptide of TbHslU1 (VDIKKFIL), only that of TbHslU2 (IDLAKYIL) could activate TbHslV ( Fig. 2A). In order to check the synergistic activation of TbHslV by both TbHslU1 and TbHslU2 peptides, we measured the activity in the presence of the mixture with both peptides. However, this did not increase the activity of TbHslV, and the level of activation correlated only with the amount of TbHslU2 peptide ( Fig. 2A). Compared with the activity of EcHslV, that of TbHslV is relatively low. To assess the activity assay result, TbHslV and peptides were set to higher concentration than the established assay conditions with E. coli enzymes. This revealed that the octapeptide, IDLAKYIL, which is an amino acid sequence derived from the C-terminal 8 residues of TbHslU2, was an activator of TbHslV activity as assayed with Z-GGL-AMC as substrate. To further confirm the action of TbHslV, we mutated the predicted catalytic threonine residue to alanine (T1A where R free is calculated for the 5% test set of reflections. mutant). As expected, this mutant enzyme does not possess any peptidase activity (Fig. 2B).

EcHslU Mutant Mimicking the C-terminal Segment of TbHslU2
Activates TbHslV for Peptide Hydrolysis-As described, EcHslU could not activate TbHslV, whereas the octapeptide IDLAKYIL, derived from the sequence of the C-terminal sequence of TbHslU2, activates the TbHslV. Therefore, we generated an EcHslU mutant containing the C-terminal 8-residue sequence of TbHslU2, termed EcHslU TbHslU2 . Interestingly, the EcHslU TbHslU2 mutant fully activates peptide hydrolysis by TbHslV (Fig. 3A). As expected, the activity clearly depends on ATP (Fig. 3B).
Next we performed the protein degradation assay in the presence of mutant HslU because folded protein substrates cannot be degraded without the ATPase activity of HslU (31). SulA, a cell division inhibitor in E. coli, is encoded by the SOS-inducible sulA gene and is a natural substrate of the HslVU complex in E. coli (35). The model protein substrate MBP-SulA is recognized by the I-domain of EcHslU and then unfolded and translocated into HslV by the ATPase, HslU (15). As shown in Fig.   FIGURE 1. Sequence alignment between HslVs and the C-terminal segment of HslUs. A, sequence alignment of HslV from T. brucei (TbHslV; UniProt ID Q383Q5), H. influenzae (HiHslV; P43772), and E. coli (EcHslV; P0A7B8). Secondary structure elements are indicated above the sequence (spring, ␣-helix; arrow, ␤-strand). The red star indicates the catalytic threonine (Thr 1 ). Blue circles indicate residues participating in polar interactions with HslU. Note that the residues marked with magenta circles are smaller substitutions that might be important for specific TbHslVTbHslU2 interaction. B, sequence alignment of the C-terminal segment in HslU from T. brucei (TbHslU1 (UniProt ID Q57VB1) and TbHslU2 (Q382V8)), H. influenzae (HiHslU; P43773), and E. coli (EcHslU; P0A6H5). The key residue in TbHslU2 for specific interaction with TbHslV is marked with a red star. Green circles indicate the residues interacting with HslV. Shading indicates residues that are identical (red) or highly conserved (yellow) in all sequences. The sequence number for T. brucei enzymes is indicated at the top of the alignment in both panels. 4A, the mixture of TbHslV and EcHslU did not trigger degradation of MBP-SulA. However, the mixture between TbHslV and EcHslU TbHslU2 hydrolyzed the substrate to a similar extent as the EcHslV-EcHslU complex. Notably, the mixture of EcHslV and EcHslU TbHslU2 also hydrolyzed the substrate (Fig.  4A, lane 5), suggesting that EcHslV is a much more robust protease than TbHslV. No degradation of MBP-SulA substrate was observed in the absence of ATP, confirming that unfolding of the substrate by HslU using ATP energy is a critical step (Fig. 4B).
Importance of Tyrosine 494 at the C-terminal Segment of TbHslU2-Next, we analyzed the sequence of the C-terminal residues of TbHslU2. Sequence alignment of the C-terminal segment of several HslUs did not give a clear indication of why only TbHslU2 was capable of activating TbHslV because the sequence at this region is highly homologous (Fig. 1B). Therefore, we synthesized various octapeptides and checked the peptidase activity systematically. Because only the peptide comprising the sequences IDLAKYIL (TbHslU2) activates TbHslV, we focused on these residues of TbHslU2 as they differ with those of EcHslU and TbHslU1, which were inactive in the assay. Amino acid residues Asp 490 , Ile 495 , and Leu 496 are strictly conserved in all peptides (all HslUs), whereas Val 489 , Leu 491 , Lys 492 , and Phe 494 in TbHslU2 are different from corresponding residues in TbHslU1 (Fig. 1B). Therefore, we generated four different mutant peptides and checked their ability to activate TbHslV (Fig. 5A). Three peptides, IDIKKFIL, VDLKKFIL, and VDIAKFIL (the mutation is underlined for clarity) displayed essentially the same activity as the wild-type TbHslU1 peptide VDIKKFIL (Fig. 5A). However, the mutant peptide VDIKKYIL showed significant activation activity.
Four residues in EcHslU (Glu 436 , Ser 439 , Arg 440 , and Phe 441 ) were divergent from those in TbHslU2 (Fig. 1B). Three peptides, IDLSRFIL, EDLARFIL, and EDLSKFIL, displayed significantly lower activity than the wild-type EcHslU peptide EDLSRFIL (Fig. 5A). The EDLSRYIL mutant fully activated the TbHslV and thus was similar to the TbHslU2 peptide. These data confirm that Tyr 494 at the C-terminal segment of TbHslU2 is a key residue for TbHslV activation. We performed the same experiments with different enzymatic reaction times and confirmed the results, as shown in Fig. 5, B and C.
The dodecameric structure of TbHslV is mediated via hexameric ring-ring interactions, which are mainly hydrophobic in nature (Ile 26 -Phe 129n , Thr 128 -Phe 129n , Ala 132 -Ile 155n , and Ala 136 -Leu 137n ). Additional polar interactions (Lys 151 -Ala 136nm /Asp 139n ) contribute to the formation of the dodecamer. These results show that polar interactions, including hydrogen bond and salt bridges, are dominant during formation of the hexameric ring, whereas hydrophobic interac-tions contribute to the donut shape of the TbHslV dodecamer (Fig. 6A). The aforementioned resides are generally well conserved with prokaryotic HslVs (Fig. 1A).
Structural Comparison of TbHslV with Other HslVs-The overall structure of TbHslV is similar to that of previously determined HslV structures from prokaryotic species, including E. coli (7,15,18), H. influenzae (16), and T. maritima (20). A superimposition of the TbHslV, EcHslV, and HiHslV structures is shown in Fig. 6B. The root mean square (r.m.s.) deviation is ϳ1.4 Å between EcHslV (Protein Data Bank code 1E94) and both crystal forms of TbHslV (Protein Data Bank codes 4HNZ (Form-I) and 4HO7 (Form-II)) with 170 matching C ␣ atoms (residues 1-27, 29 -38, 39 -118, and 120 -172). The r.m.s. deviation is ϳ1.2 Å between HiHslV (Protein Data Bank code 1G3K) and both TbHslV structures with 171 matching C ␣ An axial entrance pore at the hexameric ring is one of the conserved structural characteristics of HslV (Fig. 6A). The entrance pore of TbHslV is a circular shape with a distance of 18.2, 18.5, and 18.5 Å for Form-I and 17.2, 18.7, and 19.0 Å for Form-II, respectively (measuring the distance for three pairs of the C ␣ atom of Arg 86 on six subunits). The axial entrance pore of TbHslV is slightly smaller than that of other HslVs. The entrance pore size of TmHslV is 19.4, 20.1, and 22.0 Å. However, the entrance pore of the other HslVs has a more elliptical shape with values of 13.1, 19.1, and 25.7 Å for E. coli and 13.1, 19.1, and 25.7 Å for H. influenzae, respectively. Indeed, the size of entrance pore of HslV also varies upon complex formation with HslU (16, 20) (Fig. 7A). Loops with many basic arginine residues in the entrance pore are intrinsically flexible (15,16). Subsequently, the size of the entrance pore of HslV in the complexed state is greater than that of HslV alone and is mechanistically similar to the entrance pore of the homologous ATP-dependent protease, ClpP (49,50), and the 20 S proteasome (51,52).

DISCUSSION
ATP-dependent two-component proteases exist in all three kingdoms of life. The matching symmetry of HslVU consisting of 6-fold HslU ATPase and 6-fold HslV protease is different from that of the eukaryotic 26 S proteasome, which consists of pseudo-6-fold ATPases of the base of the 19 S regulatory particle and 7-fold 20 S proteolytic core. However, the HslV and 20 S proteasome have relatively high structural and sequence similarity, including the same catalytic N-terminal threonine residue (3), suggesting that the HslVU complex is an ancestral type of 26 S proteasome. It has been reported that the HslVU complex only exists in prokaryotes and archaea, whereas the proteasome is present in eukaryotes and archaea (3). In contrast with this hypothesis, the symmetry-mismatched two-component protease ClpXP was identified in chloroplasts and mitochondria of eukaryotes more than 2 decades ago and has been studied extensively (53)(54)(55)(56). The coexistence of the HslVU complex and proteasome in eukaryotes has been reported only recently (23, 26 -29). The HslVU complex in prokaryotes and archaea possesses a simple architecture consisting of a homododecameric HslV and two homohexameric HslUs, but archaeal and eukaryotic proteasomes display a more complicated configuration. In archaea, several proteasomal ATPases, including proteasome-activating nucleotidase and CDC48, constitute a regulatory network (57), and in eukaryotes, heterooligomeric ATPases function within the base of the 19 S regulatory particle. In contrast to prokaryotic and archaeal HslUs, the eukaryotic HslUs from T. brucei and Leishmania donovani possess two HslU homologs, HslU1 and HslU2 (28,29); this suggests several possible configurations of the T. brucei HslVU complex from a structural point of view: 1) two independent TbHslVU1 and TbHslVU2 complexes; 2) TbHslV asymmetrically capped with hexameric rings of TbHslU1 and TbHslU2, and 3) TbHslV complexed with the hetero-oligomeric TbHslU ring consisting of both TbHslU1 and TbHslU2. The latter case allows for many different combinations, such as different stoichiometries of TbHslU1 and TbHslU2 in the hexameric ring or different symmetries (3-fold, alternative arrangement; 2-fold, three consecutive arrangements) even in the 1:1 composite. As shown in our biochemical data ( Fig. 2A), the C-terminal segment of TbHslU2 successfully activates TbHslV, whereas that of TbHslU1 does not. Furthermore, TbHslU1 and TbHslU2 do not act synergistically to stimulate the protease activity of TbHslV ( Fig. 2A). These results rule out the existence of TbHslVU2 and most probably the asymmetrical capped TbHslU1-TbHslV-TbHslU2 complex. Our coexpression experiment of both TbHslU1 and TbHslU2 in E. coli did not produce any hetero-oligomers, and more importantly, TbHslV was found to form homo-oligomers and thus possesses all of the same HslU-binding pockets. The TbHslV as well as both TbHslU1 and TbHslU2 are targeted to mitochondria (28); therefore, we speculate that TbHslVU2 and TbHslU1 independently function in regulating mitochondrial DNA.
The reason why only TbHslU2 is able to activate the protease activity of TbHslV remains unclear. TbHslU2 shares high sequence similarity with TbHslU1, as well as with other prokaryotic HslUs (28). Indeed, only the C-terminal segment of HslU participates in binding with HslV (16) and can replace full-length HslU functionally (8,30). From our biochemical assay, it is evident that Tyr 494 is a key determinant of HslV activation (Fig. 5A). In order to understand the structural basis for the activation of TbHslV by TbHslU2, the crystal structure of the TbHslVU2 complex is required. Unfortunately, we were unable to obtain this crystal, but a homology model of TbHslV in complex with TbHslU2 can be built using the only available functional HslVU structure from H. influenzae (16). The C-terminal segment of TbHslU2 also shows a high degree of sequence conservation with that of HiHslU (Fig. 1B), and the structure of TbHslV is quite similar to that of HiHslV (Fig. 6B). Because there is an allosteric conformational change in HslV upon complex formation with HslU (19), we used the HslUbound HslV as a template for modeling. Therefore, the model of TbHslV complexed with C-terminal segment of TbHslU2 depends on the original complex structure. In particular, the second helix containing Arg 83 , which participates in salt bridges with the neighboring subunit, shows different structures for apo-and TbHslU2-complexed TbHslV (Fig. 7, A and C).
In the HiHslVU complex, the Arg 35 in HiHslV forms hydrogen bonds with the main chain atoms of Arg 441U and Ile 443U of HiHslU (Fig. 7D). For clarity, a "U" is used for the residues of HslU. In addition, Lys 28 and the adjacent monomer Ala 83 in HiHslV form hydrogen bonds with the main chain atoms of Leu 444U and Phe 442U of HiHslU, respectively. Arg 441U also forms a salt bridge with Glu 61 . In addition to the aforementioned interactions, two C-terminal terminal residues, Ile 443U and Leu 444U , bind tightly to the surrounding hydrophobic residues of HiHslV.
In the TbHslVU complex model, Arg 36 might form hydrogen bonds with the main chain atoms of Lys 493U and Ile 495U of TbHslU2 (Fig. 7E). The critical residue for TbHslV activation is Tyr 494U (Fig. 5A), and its equivalent residue in HiHslV is Phe 442U . Therefore, the hydroxyl moiety of TbHslU2 must play a critical role in binding. Interestingly, the hydrophobic interaction between TbHslV and TbHslU appears much weaker than that between HiHslV and HiHslV. The residues for accommodating the C-terminal tail of HiHslU, Phe 54 , Phe 57 , and Gln 114n are replaced with smaller or shorter residues (i.e. Ile 54 , Met 57 , and Thr 114n , respectively) in TbHslV. This explains why EcHslU and TbHslU1 are not able to activate TbHslV, most probably due to the weak binding. Consequently, this replacement with smaller residues may provide space for swinging over the critical tyrosine residue to achieve tighter binding in the TbHslVU2 complex (Fig. 7E). Indeed, our modeling study shows that the side chain of Tyr 494U could fit into a different rotamer position and that the hydroxyl group forms a hydrogen bond with the main chain atoms of Ala 79 (Fig. 7E). When we analyzed the sequences of several eukaryotic HslUs, many of them were found to have a tyrosine residue at the equivalent position (see the ExPASy Web site). For example, there are two potential HslUs in L. donovani, and HslU1 (E9BC50_LEIDB) and HslU2 (E9B9S7_LEIDB) have phenylalanine and tyrosine residues at the critical position, respectively. An HslV ortholog in Plasmodium falciparum also has a tyrosine residue at this site (26). Therefore, we infer that the existence of the tyrosine residue in this position can determine the selection of functional HslU molecules in the eukaryotic HslVU system. The colors for carbon atoms are the same as in A-C. The important residues for the interaction between HslV and HslU are shown as a stick model and labeled. For clarity, a subscript "U" is added for HslU residues, and a letter n is added for adjacent subunits of HslV. Oxygen and nitrogen atoms are colored red and blue, respectively.