Crystal structures of the transposon Tn5-carried bleomycin resistance determinant uncomplexed and complexed with bleomycin.

The transposon Tn5 carries a gene designated ble that confers resistance to bleomycin (Bm). In this study, we determined the x-ray crystal structures of the ble gene product, designated BLMT, uncomplexed and complexed with Bm at 1.7 and 2.5 A resolution, respectively. The structure of BLMT is a dimer with two Bm-binding pockets composed of two large concavities and two long grooves. This crystal structure of BLMT complexed with Bm gives a precise mode for binding of the antibiotic to BLMT. The conformational change of BLMT generated by binding to Bm occurs at a beta-turn composed of the residues from Gln(97) to Thr(102). Crystallographic analysis of Bm bound to BLMT shows that two thiazolium rings of the bithiazole moiety are in the trans conformation. The axial ligand, which binds a metal ion, seems to be the primary amine in the beta-aminoalanine moiety. This report, which is the first with regard to the x-ray crystal structure of Bm, shows that the bithiazole moiety of Bm is far from the metal-binding domain. That is, Bm complexed with BLMT takes a more extended form than the drug complexed with DNA.

Bleomycin (Bm), 1 an antibiotic produced by Streptomyces verticillus, is widely employed in the treatment of several neoplastic diseases including non-Hodgkin's lymphoma, squamous cell carcinomas, and testicular tumors (1,2). The Bm⅐Fe(II) complex, in conjunction with a reducing agent and oxygen, causes nucleotide sequence-specific DNA cleavage (3). It has been suggested that "activated Bm," generated by the reductive activation of oxygen by a Bm⅐Fe(II) complex, cleaves DNA (3,4). Recent studies have provided unequivocal evidence for Bm-mediated degradation of certain RNA substrates, notably transfer RNAs and tRNA precursor transcripts (5)(6)(7).
Although Bm has not been used as an antibacterial agent, almost all strains of methicillin-resistant Staphylococcus aureus, isolated in Hiroshima University Hospital, were resistant to the drug. A Bm resistance gene, designated blmS, has been cloned from chromosomal DNA isolated from methicillin-resistant S. aureus and sequenced (11). The gene product, designated BLMS, is also a Bm-binding protein (9).
The transposon Tn5, expressed in Escherichia coli, carries Bm resistance together with kanamycin and streptomycin resistances (12)(13)(14). The nucleotide sequence analysis of the Bm resistance gene, designated ble, has suggested that it encodes a protein consisting of 126 amino acids with a molecular mass of 14,058 daltons (14). In addition to its role as the Bm resistance determinant, a ble product confers survival advantage to E. coli (15,16). The amino acid sequence of the ble gene product, designated by us as BLMT (17), shares sequence homology with the BLMA protein (21%) and the Shble protein (25%). BLMT has been determined to be a Bm-binding protein (17,18).
We have determined the crystal structure of BLMA at a high resolution of 1.5 Å by the single isomorphous replacement method including the anomalous scattering effect (19). Another group has determined the crystallographic analysis of the Shble protein at a 2.3 Å resolution (20). Both groups have independently provided a model that suggests that dimeric formation of the protein generates two pockets for binding to Bm. However, because the crystallization of BLMA and the Shble protein, which are complexed with Bm, has been unsuccessful until now, the precise binding mode between the protein and Bm has not been determined.
In this study, we successfully crystallized BLMT uncomplexed and complexed with Bm; the former and latter structures were determined at 1.7 and 2.5 Å resolution, respectively. We describe the conformational differences of BLMT in the Bm-free and Bm-bound form. Although a structural model of Bm bound to an oligonucleotide, determined by two-dimensional NMR analysis, has been proposed (21,22), we report a structural model based on x-ray crystallography.

EXPERIMENTAL PROCEDURES
Crystal Preparation-BLMT, overproduced using an E. coli host vector system, was purified according to the methods described previously (17,23). For crystallization, BLMT was adjusted to a final concentration of 21 mg/ml. Crystals were obtained by vapor diffusion at 25°C using the hanging-drop method (24) with the mother liquor of 25% PEG 6000 (polyethylene glycol), 0.1 M calcium acetate, and 0.1 M sodium cacodylate at pH 7.0. These crystals belong to the orthorhombic space group C222 1 with unit cell dimensions a ϭ 81.3 Å, b ϭ 85.0 Å, and c ϭ 78.8 Å. The crystal volume per unit of mass, V M (25), is 2.42 Å 3 /Da with 2 BLMT monomers in the asymmetric unit, which corresponds to a solvent content of 49%.
For crystallization of BLMT complexed with Bm, the protein was incubated at a final concentration of 14 mg/ml for 1 h at room temperature with a 3-fold molar excess of Bm A 2 sulfate (Fig. 1). The molar ratio of Bm A 2 was suitable for complete binding to BLMT. The crystals of the BLMT⅐Bm complex were obtained by vapor diffusion at 25°C using the hanging-drop method with the mother liquor of 15% PEG 6000, 0.1 M ammonium sulfate, 0.02 M magnesium chloride, and 0.05 M MES-NaOH at pH 5.0. These crystals belong to the orthorhombic space group P2 1 2 1 2 1 with unit cell dimensions a ϭ 115.3 Å, b ϭ 117.0 Å, and c ϭ 79.9 Å. The V M is 2.14 Å 3 /Da with four 2:2 complexes (2:2 ϭ two monomeric BLMT molecules complexed with two Bm molecules) in an asymmetric unit. The value corresponds to a solvent content of 42.5%.
Data Collection and Processing-Diffraction data for the Bm-free BLMT crystals were collected with synchrotron radiation (1.0 Å wavelength) at beam line 18B of the Photon Factory, the National Laboratory for High Energy Physics, Tsukuba, Japan. The data collection was done with three crystals. Three sets of high resolution data up to 1.7 Å resolution were recorded on Fuji imaging plates (400 ϫ 800 mm) using a Sakabe's Weissenberg camera (26). The diffraction images were processed to intensity data with the program WEIS (27). The intensity data from each imaging plate were scaled to each other by the Fox and Holmes method (28) and were then merged.
The diffraction data for a crystal of BLMT complexed with Bm was collected on a Rigaku R-AXIS IIC imaging plate detector system using the mirror monochromated CuK ␣ radiation produced by a Rigaku RU-300 rotating anode generator operated at 40 kV and 100 mA. The crystal-to-detector distance was set to 95 mm. Each frame of the 1.5°c rystal oscillation was collected for 10 min. The diffraction spots in a rotation range of 90°were recorded on a total of 60 frames. The data processing was accomplished at 2.5 Å resolution with the R-AXIS IIC data processing software package. The number of reflections having adequate intensities was small; this may be because of the noncrystallographic translational symmetry causing the systematic weakness of the intensities. Details of the data collection are summarized in Table I.
Structure Determination and Refinement of BLMT Uncomplexed with Bm-The crystal structure of BLMT was solved by the molecular replacement method using the programs in X-PLOR (29). The start model was a dimeric BLMA structure previously determined at 1.5 Å resolution (19). However, the N-terminal Met 1 and Val 2 and the Cterminal Gly 121 and Glu 122 residues in BLMA were omitted from the model, together with four turn regions (Val 20 -Trp 23 , Arg 52 -Ile 57 , Trp 78 -Ala 81 , and Glu 99 -Gly 103 ).
Atomic coordinates, obtained by the molecular replacement method, were refined against the data from 10 to 4.0 Å for 20 cycles with the entire monomer as a rigid body, resulting in a crystallographic R-factor of 56.4%. Atomic refinement of the model was performed with the combination of simulated annealing (30) and conventional restrained refinement methods (31) using the program X-PLOR. In lower resolution refinement, strict noncrystallographic symmetry restraint was imposed. Noncrystallographic symmetry restraint, gradually weakened according to the progress of refinement, was finally ignored. A subset of 10% of the reflections was used to monitor the free R-factor (R free ) (32). Individual B-factors were refined by starting with a mean B-factor of 22 Å 2 estimated from the Wilson plot (33).
The refinement was started with data from 10 to 3.5 Å resolution and finished at 2.0 Å resolution. Positional parameters and refinement for individual isotropic B-factors were included in each refinement cycle. The program Xfit in the XtalView suite (34) was used for the visualization and building of the model. In the final stage of the X-PLOR refinement, R free and the R-factor were improved to 23.8 and 20%, respectively, against the 18,250 reflections with F Ͼ 2 from 10 to 2.0 Å resolution. Further refinement was performed with the data set from 5.0 to 1.7 Å resolution using the program SHELXL-97 (35). The final R-factor was 19.1% for 26,798 reflections between 5.0 and 1.7 Å resolution.
The current model contains two protein monomers in the asymmetric unit, together with 111 water molecules, two tetraethylene glycol moieties, and one calcium ion. Because Met 1 and Leu 122 -Ser 126 were invisible on the electron density due to their higher flexibility, they were excluded from the current model. All side-chain atoms of the residues from Thr 2 to Leu 121 , except Glu 98 , are included in the model. The side-chain atoms of Glu 98 were completely invisible and, above the C␤, the atoms of the Glu 98 side-chain are excluded from the model. A Luzatti plot (36) revealed that the mean coordinate error was ϳ0.2 Å. No residues except Trp 99 are in the acceptable regions of the Ramachandran plot (37,38). A peptide bond between Tyr 88 and Pro 89 is in cisconformation. The r.m.s. deviations from the ideal values are 0.007 Å in bond length, 2.0°in bond angle, and 1.23°in improper angle (Table I).
The average B-factors are 26.3 Å 2 and 39.5 Å 2 for 1892 all-protein atoms and for 138 non-protein atoms, respectively.
Structure Determination and Refinement of BLMT Complexed with Bm-Pseudo-precession photographs indicate that the reflection intensities (I (hkl)) were systematically weak if k was an odd number. When calculated using the reflections from 10.0 to 4.0 Å resolution, the second highest peak, which is 6% lower than the origin peak, was present at (x, y, z) ϭ (0, 0.5, 0) on the Patterson map. This suggests that the noncrystallographic translational symmetry parallel to the b-axis is present in the crystal. To obtain the solutions of the rotation and the translation functions by decreasing the number of independent molecules per asymmetric unit, we made modifications as follows: the cell length of b-axis was halved. The axis transformation of the crystalline lattice was performed to satisfy the international rule for orthorhombic crystals (b Ͼ a Ͼ c). In other words, we assumed that this crystal belongs to the space group P2 1 2 1 2 with the unit cell dimensions a ϭ 79.9 Å, b ϭ 115.3 Å, and c ϭ 58.5 Å having two 2:2 complexes in an asymmetric unit. Simultaneously, the data set of the structure factor amplitude (F (hkl)) was transformed as follows: when k is an odd number, the reflection was ignored. Otherwise, if k was an even number, the Miller indices of F (hkl) were transformed as (hЈkЈlЈ) ϭ (l h k/2), where (hЈkЈlЈ) are the transformed Miller indices and (hkl) are the original ones. Initially, we solved the structure by the molecular replacement method and refined the atomic parameters using the transformed structure factor amplitudes and cell parameters with the programs in X-PLOR. The Bm-free form of dimeric BLMT was provided as a search model.
The atomic coordinate, obtained by the molecular replacement method, was refined against the data from 10.0 to 4.0 Å with F Ͼ 2 for 20 cycles with the entire monomer as a rigid body, resulting in a crystallographic R-factor of 32.8%. The atomic refinement of the model was performed in the combination of the simulated annealing and the conventional restrained refinement methods, imposing strict noncrystallographic symmetry restraint. The refinement was started with data having 10.0 -3.5 Å resolution, and the upper limit was raised to 2.5 Å resolution. The noncrystallographic symmetry restraint was weakened using the R free value as an indicator. The program Xfit was used for visualizing and rebuilding the model. In the course of the refinement, because the F o Ϫ F c map showed that Bm was present in the crystal, we constructed the model of Bm into the difference electron density and included it in the subsequent refinement cycles. Several cycles of X-PLOR refinement resulted in decreasing both the R-factor and R free to 18.6 and 24.5%, respectively, for the reflections from 10.0 to 2.5 Å resolution with F Ͼ 2 .
For refinement using the original cell parameters, the refined atomic coordinate of two 2:2 complexes was transformed as (xЈ, yЈ, zЈ) ϭ (y, z, x ϩ1/4), where (x, y, z) is the atomic coordinate in the halved crystalline lattice as (xЈ, yЈ, zЈ) is in the original one. The translational term (1/4) is necessary to correct the positional shift of the crystal origin. The atomic coordinates of the adjacent molecules, which are related by the crystallographic translational symmetry parallel to c-axis in the halved crystalline lattice, were transformed according to the above formula. The atomic coordinates of four 2:2 complexes were refined against the data between 10.0 and 3.5 Å with F Ͼ 2 for 20 cycles considering the entire 2:2 complex as a rigid body, resulting in a crystallographic R-factor of 21.7%. Because the number of reflections with adequate intensities was smaller when compared with the number of refined parameters, we paid attention to the following points: the noncrystallographic symmetry restraint imposing on the proteins and Bm was strongly set and decreased according to the R free value. After the bulk solvent correction (39) was done, reflections at low resolution together with weak reflections with F Ͻ 2 were used for the refinement. Further cycles of atomic refinement yielded the current model. The R-factor and the R free were improved to 22.1 and 30.2%, respectively, against all reflections from 30.0 to 2.5 Å resolution. When limited to the reflections with F Ͼ 2 , the final R-factor and the R free were 19.0 and 26.4%, respectively.
In the present study, we prepared the original topology and parameter files of the Bm molecule. The metal-binding domain in the Bm A 2 molecule is composed of the ␤-aminoalanine, pyrimidinyl propionamide, and ␤-hydroxyhistidine moieties. The binding domain for DNA is composed of the ␥-aminopropyl dimethylsulphonium and bithiazole moieties (Fig. 1). The parameters used for construction of a three-dimensional model of the former and latter moieties were obtained from the x-ray crystal structures of the P-3A⅐Cu(II) complex (40) and 3-(2Ј-phenyl-2,4Ј-bithiazole-4-carboxamide)propyl dimethylsulfonium iodide (41), respectively. The parameters for threonine and methyl valerate in the linker moiety and for gulose and mannose in the sugar moiety of the Bm molecule were prepared using the energy minimization structure, which is calculated by the program CAChe.
The current model contains 8 monomeric BLMT molecules, 8 Bm molecules, and 48 water molecules in the asymmetric unit. The monomeric BLMT molecule in the current model consists of the Thr 2 -Glu 120 residues, because both terminal residues, Met 1 and Glu 121 -Ser 126 , were invisible in the electron density. All of the side-chain atoms of the Thr 2 -Glu 120 residues are included in the model. In the electron density, the ␥-aminopropyl dimethylsulphonium and the ␤-aminoalanine moieties of Bm are almost invisible and poorly defined, respectively. The mean coordinate error was estimated from a Luzzati plot to be 0.32 Å. The Ramachandran plot indicates that the backbone torsion angles for 67.9% of the nonglycine residues are in the favor region, whereas those for 95.9% are in the acceptable region. Most of the remaining residues, which fall outside of the acceptable region, have poorly defined electron densities and may have large errors in the parameters. The r.m.s. deviations from the ideal values are 0.007 Å in bond length, 1.2°in bond angle, and 0.57°in improper angle (Table I). The average B-factors are 28.1 Å 2 for 7544 all-protein atoms and 44.2 Å 2 for the 768 atoms of the Bm molecule.
Measurement of Dissociation Constant of BLMT for Bm-Bm A 2 sulfate was added at a range of 0.05 to 0.5 M to 2 ml of 1 mM Tris-HCl (pH 7.5) containing 0.4 M BLMT. Experiments were done at 25°C using a spectrofluorophotometer (model RF-5000, Simadz, Japan) under the condition of excitation ϭ 280 nm, emission ϭ 350 nm, slit width at excitation ϭ 10 nm, and slit width at emission ϭ 10 nm. To estimate the concentration of protein contained in solution, the molar absorptivity (⑀ ϭ 19,800), calculated from the number of Trp and Tyr residues in the BLMT molecule, was employed. The dissociation constant of BLMT for Bm was calculated as described previously (19).
Two monomeric BLMT molecules are related by a noncrys-tallographic 2-fold axis (Fig. 3A). The ␤1-strand interacts with the partner's ␤6-strand, suggesting that the former strand plays a key role for the dimeric formation. The topology of BLMT is almost the same as BLMA (19) and the Shble protein (20). The dimer formation of BLMT, generated by the alternate arm exchange of two monomeric BLMT molecules, results in two large concavities and two long grooves. Although a noncrystallographic symmetry restraint is not imposed in the final refinement, one monomeric BLMT molecule is almost the same as the partner's monomer. The superposition of main-chain atoms on each BLMT monomer is shown in Fig. 3B. The r.m.s. positional differences between two monomeric BLMT molecules are 0.41 Å for main-chain atoms and 0.79 Å for all-protein atoms. Structural deviations occur at two turn regions, Ala 49 -Ser 58 and Leu 96 -Thr 102 , and at a long loop composed of the Gln 82 -Gly 87 residues. The conformational deviations may be generated because of the flexibility of these regions.
The overall structure of BLMT complexed with Bm is almost the same as that of the Bm-free BLMT (Fig. 4). The current model shows with accuracy that two Bm molecules bind to two Bm-binding pockets formed by the alternate arm exchange of two monomeric BLMT molecules.

Binding Mode between Bleomycin and BLMT
The bithiazole moiety of Bm is inserted into the long groove running along the dimer interface (Fig. 5A). The first thiazolium ring, which is interacted by the hydrophobic effect with the Phe 30B benzene, is also interacted by the polar effect with the Arg 65A guanidino group (Fig. 5B). The second thiazolium ring is tightly stacked with two indole rings from Trp 35B and Trp 99A (Fig. 5B). This stacking effect contributes to stabilization of the Bm molecule.
The Trp 35B indole ring is interacted by a hydrogen bond with the Glu 46B carboxylate. The Glu 46B carboxylate, interacted by two salt bridges with the Arg 65A guanidino group, is also stabilized by a hydrogen bond with the Thr 6B hydroxyl oxygen (Fig. 5B). That is, a hydrogen bonding network is extended from Trp 35B to Arg 65A . In addition, the Arg 65A and Trp 35B are interacted with the first and second thiazolium rings, respectively.
The large concavity is formed mainly by the residues from one monomer. One side of the concavity has the hydrophobic residues. The linker, metal-binding, and sugar domains of Bm are buried in the large concavity (Fig. 5A) and stabilized by a large number of hydrogen bonds to the protein atoms.
In the metal-binding domain of Bm, an amino group attached  to the pyrimidine ring is interacted by a hydrogen bond with two backbone carbonyl oxygens of Phe 60A and Gly 111A (Fig. 6). The carbonyl oxygen of the pyrimidinyl propionamide moiety of Bm forms a hydrogen bond with the Arg 90A guanidino group (Fig. 6). The terminal amide group of the pyrimidinyl propionamide moiety forms three hydrogen bonds: its oxygen atom with the Arg 115A guanidino group, its nitrogen atom with the Ser 61A hydroxyl oxygen, and the Trp 59A carbonyl oxygen (Fig.  6). However, ␤-aminoalanine and ␤-hydroxyhistidine for the metal binding make no polar interactions with the protein atoms. Because the ␤-aminoalanine moiety is poorly defined in the electron density, the conformation may be unstable. However, its amide group orients parallel to the Trp 59A indole ring and seems to be stacked by the apolar interaction. In fact, the Trp 59A side-chain is flexible in the Bm-free form but rigid in the complexed form because the side-chain is stacked with the imidazole rings of His 50B and the amide group of the ␤-aminoalanine moiety of Bm. Only mannose, but not gulose, in the sugar moiety of Bm is involved in the binding to the protein. The carbamoyl group, attached to the 3Ј hydroxyl oxygen of mannose, makes two hydrogen bonds: its nitrogen atom with the carbonyl oxygens of Pro 55A and Ser 58A (Fig. 6). The 2Ј and 4Ј hydroxyl groups form a hydrogen bond with the Leu 56A carbonyl oxygen and the Arg 90B guanidino group, respectively. The 6Ј hydroxyl group interacts by the hydrogen bond with the Arg 90B guanidino, Ser 85B carbonyl, and Gly 87B amide groups (Fig. 6).
The interaction between the protein and the linker domain of Bm is weak. Only one hydrogen bond is formed between the Arg 115A guanidino group and the carbonyl oxygen of the threonine moiety of Bm (Fig. 6). This occurrence is likely to be related with the flexibility of the domain.
The terminal ␥-aminopropyl dimethylsulphonium moiety of Bm A 2 , having the positive charge, is not well defined in the electron density. The end of the long groove, which is the expected binding site for the ␥-aminopropyl dimethylsulphonium moiety, has negatively charged residues Asp 3 , Asp 42 , Asp 67 , and Glu 120 . These residues are located around the positive charge of the ␥-aminopropyl dimethylsulphonium moiety. However, no electrostatic interactions are observed between this moiety and the negatively charged residues, suggesting that the negatively charged residues are necessary for the recognition of the ligand rather than for stabilization. Fig. 7A shows the superposition of the BLMT⅐Bm structure on the Bm-free structure. The positional r.m.s. differences between two structures are 0.53 and 0.77 Å for the main-chain atoms and all-protein atoms, respectively. The marked positional deviation occurs at the residues located on the loop and turn regions, such as His 50 -Ala 57 , Glu 83 -Gly 87 , and Gln 97 -Thr 102 . In particular, the Gln 97 -Thr 102 residues, composed of a ␤-turn between the ␤7and ␤8-strands, move about 2 Å toward the partner's monomer. The conformational change enables the Trp 99 indole ring of one monomer to approach the Trp 35 indole ring of the partner's monomer. The conformational approach results in the formation of stable structure of the long groove and strong intercalating interaction with the bithiazole moiety of Bm. However, the corresponding region of the Bm-free BLMT appears to be highly flexible with large B-factors (Fig. 7B).

Comparison with Bm-binding Proteins from Other Microbial Sources
Overall Structure-The amino acid sequence homology between BLMA and the Shble protein is ϳ60%, whereas that between BLMT and BLMA, or between BLMT and the Shble protein is only 21-25%. Nevertheless, the superposition of the main-chain atoms of BLMT on that of BLMA or the Shble protein shows that the overall structure of BLMT is almost the same as those of BLMA and the Shble protein (Fig. 8A). The r.m.s. positional differences for 114 residues of the core region composed of ␤-sheets are 1.25 Å between BLMT and BLMA and 1.35 Å between BLMT and the Shble protein. The marked structural deviation found among three Bm-binding proteins occurs away from the gravity center of dimeric molecules.
The conformation of a long loop connecting ␣2-helix and the ␤7-strand in BLMT is more stable than that in BLMA or the Shble protein. The stability is likely to be generated by the deletion of one amino acid in this region (Fig. 8B). Otherwise, the turn region between the ␤7and ␤8-strands of BLMT is highly flexible. The flexibility may be generated by the insertion of an amino acid into the long loop (Fig. 8B).
Substrate Specificity- Table II lists the dissociation constants (K d ) to Bm and phleomycin (Phm) for BLMT and other Bm-binding proteins. Phm is an analog of Bm and has a thiazolinylthiazole instead of a bithiazole moiety in Bm; that is, the second thiazolium ring in Phm is reduced and nonplanar. We have observed that the affinity between BLMT and Phm is stronger than between BLMA and Phm. Although the K d value to Phm for the Shble protein has not been determined yet, there is a report that the minimum inhibitory concentration of Phm to E. coli expressing the Shble protein is much lower than that of Bm (20). This suggests that the binding affinity of the Shble protein to Phm is lower than to Bm. The difference of the dissociation constant between Bm and Phm can be explained as follows. The turn region of BLMT composed of the Gln 97 -Thr 102 residues, which is the Bm-binding domain, is more flexible than those of BLMA and the Shble protein. The flexibility of the binding domain in BLMT is generated by the insertion of an

Structure of Tn5 Bleomycin-binding Protein⅐Bleomycin Complex
additional amino acid into the turn region (Fig. 8B). The flexibility might decrease the substrate specificity for binding.
Bm-binding Sites-Aromatic residues corresponding to Phe 30 and Trp 35 of BLMT, which interact with the bithiazole moiety of Bm, are present in BLMA, BLMS, and the Shble protein (Fig. 8B), suggesting that the aromatic rings are interacted by the hydrophobic effect with the bithiazole moiety. An amino acid corresponding to Trp 99 of BLMT, which interacts with the bithiazole moiety of Bm, is also present in BLMS and the Shble protein. The Trp residue of these proteins is likely to interact with the bithiazole moiety. BLMA is replaced by Ala instead of Trp 99 . A crystallographic analysis of BLMA complexed with Bm showed that Pro 101 interacts with the bithiazole moiety. 2 The negatively charged residues of the BLMT⅐Bm complex are located around the positive charge of the ␥-aminopropyl dimethylsulphonium moiety of Bm A 2 , but no electrostatic interactions are found between the protein and Bm. The distribution of the negatively charged residues in BLMT (Fig. 8B) resembles those in other Bm-binding proteins (19,20). These negatively charged residues might be helpful in the recognition of the ligand.

Structure of Bleomycin
Despite the powerful investigation of the x-ray crystal structure of Bm for the last two decades, the structure has not yet been determined. Therefore, the controversial and elusive points on the structure of the Bm molecule must be resolved; for example, it is not yet clear whether the two thiazolium rings in the bithiazole moiety of the Bm molecule are in cis-or trans-conformation. The crystallographic study of bithiazole derivatives has suggested that the two thiazolium rings may be in trans-conformation (41,42). Some groups have determined that the solution structure of bithiazole moiety bound to DNA and suggested that the cis-conformation is favored to account for the upfield shift of the chemical shift of the bithiazolium ring protons (43,44). In the present study, we tried to model both cis-and trans-type bithiazole moiety. From the electron density map obtained in this study, we conclude that the two thiazolium rings are in trans-conformation.
The ligands of the Bm molecule for the metal ion and its chirality also remain ambiguous. Most investigators agree on the fact that the equatorial ligands are the secondary amine of the ␤-aminoalanine moiety, the amide nitrogen of the ␤-hydroxyhistidine moiety, and the nitrogens from the pyrimidine and the imidazole rings. However, three possibilities are emphasized for the axial ligand: (i) the primary amine of the ␤-aminoalanine moiety (45)(46)(47); (ii) the carbamoyl nitrogen of the mannose moiety; and (iii) the primary amine of the ␤-aminoalanine moiety and the carbamoyl nitrogen of the mannose moiety (48 -50). Although our model of BLMT complexed with Bm does not contain the metal ion, we suggest that the primary amine of the ␤-aminoalanine moiety is suitable as an axial ligand for the metal ion. In the Bm molecule, while the car-

Structure of Tn5 Bleomycin-binding Protein⅐Bleomycin Complex
bamoyl group of the mannose moiety takes a stable conformation by forming the hydrogen bonds to the protein atoms, the conformation of ␤-aminoalanine moiety is unstable because of the lack of polar interactions with the protein atoms. Furthermore, the primary amine of the ␤-aminoalanine is just positioned over the putative equatorial plane for the metal ion with the same chirality as the propositions (45)(46)(47). Judging from these observations, the primary amine of the ␤-aminoalanine moiety may be an axial ligand.
The most interesting aspect in the research of the structure of Bm is the interpretation of the DNA-cleavage activity of Bm. Until now, models of Co(III)⅐Bm A 2 complexed with DNA have been built by NMR analysis (21,22). In the model, the bithiazole moiety is inserted into the space between two base pairs and stacked with the pyridine and pyrimidine rings. The positively charged terminal amine of Bm is buried in the major groove of DNA and is interacted with the negatively charged backbone phosphate. The linker and metal-binding domains of Bm, located in the minor groove of DNA, are stabilized by several hydrogen bonds with DNA including base pair-specific ones.
The binding mode of Bm complexed with BLMT may be similar to that with DNA for the following reasons. First, the bithiazole moiety is intercalated with the aromatic rings in both cases of DNA and BLMT. Second, the linker and metalbinding domains are buried in the minor groove of DNA or the large concavity of BLMT and stabilized by a number of hydrogen bonds. Finally, the positive charge of the terminal amine of Bm may be necessary for the electrostatic interaction with DNA or BLMT.
A model of the DNA⅐Bm complex shows that Bm exhibits a more compact form, that is, the bithiazole moiety folds back toward the metal-binding domain (21,22). However, x-ray analysis of Bm complexed with BLMT shows that the Bm molecule has a more extended form, that is, the bithiazole moiety is far from the metal-binding domain.