Enacyloxin IIa Pinpoints a Binding Pocket of Elongation Factor Tu for Development of Novel Antibiotics*

Elongation factor (EF-) Tu·GTP is the carrier of aminoacyl-tRNA to the programmed ribosome. Enacyloxin IIa inhibits bacterial protein synthesis by hindering the release of EF-Tu·GDP from the ribosome. The crystal structure of the Escherichia coli EF-Tu·guanylyl iminodiphosphate (GDPNP)·enacyloxin IIa complex at 2.3 Å resolution presented here reveals the location of the antibiotic at the interface of domains 1 and 3. The binding site overlaps that of kirromycin, an antibiotic with a structure that is unrelated to enacyloxin IIa but that also inhibits EF-Tu·GDP release. As one of the major differences, the enacyloxin IIa tail borders a hydrophobic pocket that is occupied by the longer tail of kirromycin, explaining the higher binding affinity of the latter. EF-Tu·GDPNP·enacyloxin IIa shows a disordered effector region that in the Phe-tRNAPhe·EF-Tu (Thermus aquaticus)·GDPNP·enacyloxin IIa complex, solved at 3.1 Å resolution, is stabilized by the interaction with tRNA. This work clarifies the structural background of the action of enacyloxin IIa and compares its properties with those of kirromycin, opening new perspectives for structure-guided design of novel antibiotics.

EF-Tu⅐GDPNP⅐enacyloxin IIa crystals were mounted directly from mother liquor and flash-cooled in a 100 K cryostream (Oxford Cryosystems). X-ray crystallographic data extending to 2.3 Å resolution were collected from a single crystal at beam line I711 (MAX-Lab). Phe-tRNA Phe ⅐EF-Tu⅐GDPNP⅐enacyloxin IIa crystals were flash-cooled in a cryoprotecting buffer consisting of 2.6 M (NH 4 ) 2 SO 4 , 22% trehalose, 30 mM Tris-MES, pH 6.5, 10 mM MgCl 2 , 0.5 mM DTT, and 50 mM enacyloxin IIa, and data were collected at 100 K at beam line X13 (EMBL/ DESY). Diffraction data were integrated and scaled with the HKL package (10), and structure factor amplitudes were calculated using Truncate of the CCP4 package (11). Table 1 shows the data statistics. The tRNA-bound complex (Table 1) suffered from strong anisotropy in the diffraction pattern, compromising data quality and subsequent refinement procedures.
Structure Determination and Refinement-Structures were determined by molecular replacement using AMoRe (12) and search models derived from a kirromycin-bound complex (Protein Data Bank entry 1OB2). 6 Two molecules of the EF-Tu⅐GDPNP⅐enacyloxin IIa complex related by a noncrystallographic 2-fold rotation axis (NCS) were located. Model phases were refined and extended by density modification using averaging, solvent flattening, and histogram matching in the program DM at 30 -2.3 Å resolution (13). The resulting map was practically unbiased and showed clear density for enacyloxin IIa (Fig. 1B). Model building was performed in O (14), and an assigned C13-R, C14-S configuration for enacyloxin IIa provided the best fit. The model derived from the DM map was refined using crystallography NMR software (CNS) (15) initially at 5-2.3 Å resolution, enforcing strict NCS, and later at 30 -2.3 Å resolution, using restrained NCS and bulk solvent correction. The A -weighted 2F o Ϫ F c and F o Ϫ F c electron density maps were used in later rounds of model rebuilding. The final model comprised amino acid residues 9 -40 and 55-393 for chain A and residues 6 -41 and 57-393 for chain B as well as complete models of GDPNP⅐Mg 2ϩ and enacyloxin IIa for both complexes. A total of 198 water molecules identified at hydrogen-bonding positions were included in the final model, and the final R-factor and R free factor of the model are 0.231 and 0.270, respectively, at 30 -2.3 Å resolution (all data).
The EF-Tu⅐GDPNP⅐enacyloxin IIa⅐Phe-tRNA Phe structure contained three molecules in the asymmetric unit related by 3-fold NCS as observed also for the ternary complex without antibiotic (16) and with kirromycin. 6 The complex was refined at 30 -3.1 Å resolution as a single protomer with strict NCS throughout. The final model comprises a complete Phe-tRNA Phe molecule, EF-Tu residues 9 -405, GDPNP⅐Mg 2ϩ , and enacyloxin IIa. The final R-factor and R free factor converged at 0.284 and 0.296, respectively, at 30 -3.1 Å resolution (all data).
The figures were prepared using PyMol (33).

Enacyloxin IIa Binding Site on EF-Tu-Enacyloxin
IIa (for its structure, see Fig. 2A) is active against Gram-positive and Gram-negative bacteria, slightly active against fungi, but inactive against yeasts (17)(18)(19). It stabilizes the EF-Tu⅐GTP complex (7) and enables EF-Tu⅐GDP to interact with aa-tRNA and ribosomes (8). Crystallographic analysis of EF-Tu⅐GDPNP⅐enacyloxin IIa locates the antibiotic at the "back" of the domain 1-3 interface (Fig. 1A), where a well ordered electron density delineates its boomerang-like structure (Fig. 1B). As a consequence of its insertion the interface is widened, and domain 1 is pushed upward and tilted forward (Fig. 1C). The antibiotic also reorients the side chain Strong anisotropy in the C2 form caused a low completeness in the higher resolution bins.
where R free is the R-factor calculated for a subset of ϳ1200 reflections excluded from the refinement throughout. d Percentages of amino acid residues occupying the most favored and disallowed regions, respectively, of the Ramachandran plot according to PROCHECK (31).
of several interface residues, breaking the hydrogen bonds Gln 124 N⑀2-Phe 374 O and Tyr 160 OH-Glu 315 O⑀. Thus, it binds by an induced-fit mechanism. At Ͻ3.8 Å it contacts 16 amino acids (7 from domain 1 and 9 from domain 3) via 6 hydrogen bonds and extensive van der Waals and hydrophobic interactions (Fig. 2, A, C, E, and G). A salt bridge links the carboxyl group of the hexane ring with the side chain N of Lys 313 (Fig.  2, A and C). It is noteworthy that all residues in the text are numbered according to EF-Tu(Ec), even when referring to EF-Tu(Tt)) or EF-Tu(Ta), with the homologous residue number in italics between parentheses. For more details of the enacyloxin IIa binding site, see Table 2.
Similarities to and Divergences from the Kirromycin Binding Site-In this work kirromycin is often used as synonym for methylkirromycin ( Fig. 2B), because it is the prototype and most studied member of this family and displays the same effect on EF-Tu (20). From the threedimensional model of the GTP-like EF-Tu(Tt)⅐GDP⅐kirromycin com- . C compares enacyloxin IIa-bound EF-Tu to the GDPNP-bound form of EF-Tu without antibiotic. The view is from the right side and slightly from below, compared with that in A. The insertion of enacyloxin IIa in the domain 1-3 interface at the "back" causes domain 1 to be pushed up and tilted slightly "forward" as indicated by the arrows. D shows that, compared with enacyloxin IIa, kirromycin causes in addition to the upward push and forward tilt of domain 1, an additional lateral displacement, roughly along the axis of helix A, by about half a turn. FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 Note that the heads of the two antibiotics point in opposite directions; that of enacyloxin points toward the viewer, whereas that of kirromycin points away (into the EF-Tu). Also note that the tail of kirromycin points away into the interior of EF-Tu. This latter point is illustrated plex (6) the antibiotics binding site at the domain 1-3 interface shows striking similarities to that of the structurally unrelated enacyloxin IIa (Figs. 1D and 2, C and D) but also marked divergences. The surface area buried by enacyloxin IIa (ϳ1200 Å 2 ) is similar to that buried by kirromycin bound to EF-Tu⅐GDP or Phe-tRNA Phe ⅐EF-Tu⅐GDPNP. 6 The bulkier kirromycin widens the domain 1-3 interface of EF-Tu slightly more than enacyloxin IIa (by 15.4°versus 13.7°) (cf.  Table 2). Enacyloxin IIa makes hydrogen bonds with Arg 123(124) , Gln 124(125) , and Tyr 160(161) , as does kirromycin, although enacyloxin IIa only makes hydrogen bonds with the Arg 123 main chain O, whereas kirromycin does so with the Arg 123(124) side chain (Fig. 2,  A-D). Enacyloxin IIa has two more hydrogen bonds to main chain residues, and kirromycin lacks the salt bridge to Lys 313(325) . In the enacyloxin IIa complex, the amino acid residues interacting with the antibiotic at Ͻ3.8 Å are interconnected with 4 hydrogen bonds and in the kirromycin complex with 8 hydrogen bonds. For more details of the two antibiotics binding sites, see Table 2.

Binding of Enacyloxin IIa to EF-Tu
The hydrophobic contacts of enacyloxin IIa with domain 1 residues show important differences from those of kirromycin. The enacyloxin IIa polyenic chain is loosely bound at a patch with Gln 159 , Tyr 160 , Asp 161 , and Arg 123 on the one side and Leu 120 , Arg 123 , and Gln 124 on the other side, steering clear of the side chain of the latter (Fig. 2C). On the other hand, the chain of kirromycin interacts closely with both Gln 124(125) and Leu 120(121) , whereas the central pentose tightly fills the groove at the domain 1-3 interface, because of its hydrogen bonds with Arg 123(124) (Fig. 2, D, F, and H). The "head" of enacyloxin IIa touches the tip of the loop formed by domain 3 residues 313-316, whereas that of kirromycin points the other way, wrapping around Leu 120(121) and Tyr 160(161) and displacing the Glu 117(118) side chain. Furthermore, the "tail" of kirromycin fills the hydrophobic pocket between Gln 124(125) and Ile 92(93) , whereas enacyloxin IIa passes along the outside of the cavity (Fig. 2, G  and H). This is in line with the weaker binding of enacyloxin IIa, which is released from EF-Tu on short gel filtration columns, whereas kirromycin is not (see Ref. 7 and the representative experiment in Fig. 3). With enacyloxin IIa the presence of aa-tRNA accelerates this effect. This is important because in the cell most EF-Tu⅐GTP is bound to aa-tRNA. The results of this experiments demonstrates that the binding affinity of enacyloxin IIa for EF-Tu is much lower that that of kirromycin, as expected from the properties of their binding to EF-Tu. Under the chosen conditions a 50% release of enacyloxin IIa from EF-Tu in ϳ20 min can be estimated, whereas that of kirromycin is negligible. The determination of the dissociation constant of the two EF-Tu⅐antibiotic complexes was not carried out because of the very low water solubility of enacyloxin IIa and kirromycin and their tendency to induce aggregation of EF-Tu. The domain 3 residues contacting both antibiotics form a more or less flat surface, with bumps that are in both cases tightly complementary to the antibiotic. Arg 373(385) "embraces" the active core (Fig.  2, C, D, G, and H).
Overall and Specific Changes in EF-Tu Domains and Phe-tRNA Phe Complex-In EF-Tu⅐GDPNP⅐ enacyloxin IIa the overall domain arrangement resembles that of EF-Tu⅐GDPNP in the absence of the antibiotic (Fig. 1C). In the enacyloxin IIa complex the C terminus of

TABLE 2 Connections between enacyloxin IIa and its EF-Tu binding site and hydrogen bonds interconnecting residues at <3.8 Å from the antibiotic
A comparison with kirromycin is shown. All residues are numbered according to EF-Tu(Ec) even when referring to EF-Tu(Tt) as in the methylkirromycin complex. The homologous T. thermophilus residue number is shown in parentheses and italics. Bold letters indicate common amino acid residues. more clearly in H. In E-H, the EF-Tu is shown as a molecular surface with selected elements identified by labels. The color scheme is again red for domain 1 and blue for domain 3. In E and F, the complex relationships between kirromycin and Gln 124(125) (green area) with several contacts is compared with that of enacyloxin IIa. For a close comparison, E and F also outline the overlap with kirromycin and enacyloxin IIa, respectively. The green area in G and H is the hydrophobic surface of atoms within 4 Å of the kirromycin tail (or the equivalent atoms in G). Clearly the tail of enacyloxin IIa borders this pocket, whereas that of kirromycin occupies it. Note that in C and D, domains 1 are in the exact same position, whereas domains 3 are pushed down and to the left as a result of the greater bulk of kirromycin. This larger gap between the domains can be recognized in H on the left side (cf. G).
helix C (residues 89 -93) is slightly pushed up, whereas helix B is pulled roughly along its axis toward the tail of the antibiotic. As a consequence the effector region, residues 41-65, that in EF-Tu⅐GDPNP contacts this helix is destabilized and changes conformation. Its N-terminal part (residues 41-50) shows no electron density, although in the residual part (Switch 1 region) the weak electron density reveals an unwound a-helix AЉ collapsed into domain 2 (Fig. 1A). Arg 58(59) and Thr 61(62) , two conserved residues important for GTPase activity and interaction with the ribosome (21,22), are moved away by ϳ15 Å toward the vacant binding pocket for the aminoacyl residue of tRNA. Fig. 1D illustrates the superimposition of EF-Tu⅐GDPNP⅐enacyloxin IIa and EF-Tu⅐GDP⅐kirromycin. Also the kirromycin complex shows a disordered effector region (6). For the changes in the Switch 1 region of the two antibiotic complexes, see also Fig. 6. Most important, on binding of Phe-tRNA Phe to EF-Tu⅐GDPNP⅐enacyloxin IIa the disordered effector region is restructured by the interaction of the ␣-helix AЉ with tRNA (Fig. 4). The acceptor stem and part of the CCA-end of tRNA are tugged by the effector region to follow the displacement of domain 1 induced by the antibiotic binding, whereas the remaining tRNA regions follow domains    EF-Tu⅐[␥-32 P]GTP was preformed by incubating EF-Tu⅐GDP (200 -400 pmol) for 20 min at 30°C in a 65-l solution of 50 mM Tris-HCl, pH 7.5, an excess (5-20-fold) of [␥-32 P]GTP (specific activity, 500 -4000 cpm/pmol) over EF-Tu, 10 mM MgCl 2 , 2 mM phosphoenolpyruvate, 50 g/ml pyruvate kinase, and 1 mM ATP. Thereafter, LiCl (E), NaCl (Ⅺ), KCl (‚), or NH 4 Cl (छ), to a final concentration as indicated plus 50 M enacyloxin IIa (A) or kirromycin (B), was added to aliquots of the reaction mixture kept at 0°C. GTP hydrolysis was started by a temperature shift to 37°C. Several samples were withdrawn in a period of time (15-60 min) in which the GTPase rate was linear, and the GTPase activity was determined for each salt concentration kinetically as 32 P i liberation after charcoal treatment (32). The background of the intrinsic GTPase activity of EF-Tu in the absence of antibiotics was subtracted. Data points represent an average over three experiments, and error bars indicate S.D.
2 and 3. As a result, the acceptor stem twists relative to the T-stem and displays a unique conformation intermediary between free tRNA and tRNA in the ternary complex (Fig. 4). The phenylalanyl group has a reduced electronic density, indicating a defective accommodation (Fig.  5) that is reflected in the decreased protection by EF-Tu⅐GTP against the spontaneous deacylation of Phe-tRNA Phe , which is more pronounced than in the case of kirromycin (see supplemental Fig. 1). The Switch 2 region of the enacyloxin IIa complex shows no marked changes. The side chain of His 84(85) , a residue important for the GTPase activity dependent on EF-Tu (23)(24)(25), points away from the ␥-phosphate of the nucleotide similarly to the native conformation (Fig. 6). In contrast, in the kirromycin complex it is turned toward the phosphate groups of GDP. Enacyloxin IIa enhances the intrinsic GTPase activity of EF-Tu but much more weakly than kirromycin ( Fig. 7; note the difference in scale between panels A and B). Interestingly, with enacyloxin IIa the different ions show similar concentration dependence, the "soft" NH 4 ϩ being however the strongest stimulator, whereas with kirromycin EF-Tu displays a GTPase increase as a function not only of the increasing concentration of "hard" monovalent ions but also of their charge density.

DISCUSSION
This work describes for the first time the structure of the EF-Tu⅐enacyloxin IIa complex, defining the antibiotic binding site on EF-Tu, which in the domains 1-3 interface overlaps that of kirromycin. There is strong evidence that despite a number of functional differences both compounds inhibit protein synthesis by inducing a unique GTPlike conformation of EF-Tu⅐GDP (8). Accordingly, the overall structures of EF-Tu⅐GDPNP⅐enacyloxin IIa and EF-Tu⅐GDP⅐ kirromycin, and the binding sites of the two antibiotics share marked similarities. Nonetheless there are important, specific divergences, such as the less complex contact of enacyloxin IIa with Gln 124(125) and the behavior of the short tail of enacyloxin IIa, which borders an empty hydrophobic pocket, whereas the longer tail of kirromycin fits into this pocket. This, we believe, represents the main reason for the lower binding affinity of enacyloxin IIa to EF-Tu. Related to this aspect is the differential effect caused by substituting the bordering Ala 375(387) with the bulkier Val that fills this pocket. EF-Tu becomes 3-fold more sensitive to enacyloxin IIa (8) but ϳ300 times more resistant to kirromycin (26). The insertion of the Val 375 side chain stabilizes the short tail of enacyloxin IIa by extending the contacting hydrophobic surface (Fig. 8A), whereas it inhibits the fitting of the longer tail of kirromycin by steric hindrance (Fig. 8B). Fig.   8C illustrates the overlap of enacyloxin IIa and kirromycin, underlining their overall structural divergences and similarities; it also indicates the active core of kirromycin containing the tail half-molecule as deduced from the functional analysis of fragments and chemical modifications of the antibiotic (27)(28)(29). Our observations hint at new approaches for a structure-guided antibiotic design on the basis of the hybridization of enacyloxin and kirromycin structural components, with the aim of obtaining novel EF-Tu binding compounds with specific properties. From our results, combination of the head moiety of enacyloxin IIa with the tail moiety of kirromycin should increase the binding affinity of the resulting compound. The binding efficiency of an antibiotic can affect the degree of the antibacterial effect and reduce the onset of resistance by microorganisms. One can extend this approach by creating hybrids between structural parts of the two antibiotics and other cyclic compounds (e.g. sugars, lactones, macrolides, etc.). By taking into account the common properties of the two antibiotics binding sites, one could combine any structural moieties making hydrogen bonds with amino acids around the head of the antibiotic, e.g. Tyr 160(161) with compounds bonding residues around the tail, such as Glu 124(125) and Lys 313(325) .
Another interesting result of this work is the ability of aa-tRNA to revert the disorder of the effector loop caused by the binding of the antibiotic. This effect is associated with a unique distortion of the acceptor stem of tRNA. This, however, does not impair the EF-Tu-mediated binding of aa-tRNA to the programmed ribosome, which in the presence of enacyloxin IIa is even improved (7). A possible explanation for this effect can by derived from the observation that the similarly distorted tRNA bound to EF-Tu⅐GDPNP⅐kirromycin shares common features with the conformation of the ribosome-bound EF-Tu⅐GDPNP⅐kirromycin complex. 6 Concerning the high mobility of the effector region, a plausible mechanism can be found in the distortion of helices B and C induced by the accommodation of the antibiotic in the domain-1-3 interface. The signal arising from the binding of the antibiotic can thus be transferred to other areas of EF-Tu.
It is not clear why EF-Tu⅐enacyloxin IIa shows a lower intrinsic GTPase activity than EF-Tu⅐kirromycin. Site-directed mutagenesis did not prove the involvement of His 84 side chain in a nucleophilic attack on the ␥-phosphate; however, it supported an indirect key role of this residue in both the intrinsic and the very fast GTP hydrolysis evoked by programmed ribosomes (23,24). To date the most convincing hypothesis about the mechanism of the low intrinsic EF-Tu GTPase activity is "substrate-assisted" GTP hydrolysis (30). The level of intrinsic GTPase activity of EF-Tu likely depends on selective local pH effects influencing FIGURE 8. Structural divergences between enacyloxin IIa and kirromycin and some specific effects. A and B show the structural background of the effect of the A375V mutation on the sensitivity of EF-Tu toward enacyloxin IIa and kirromycin. The panels show close-up views in the same orientation of the enacyloxin IIa (A) and the kirromycin (B) complex of EF-Tu. Ala 375 has been substituted in silico by Val (gray). The transparent van der Waals sphere around the ␤-carbon indicates the space occupied by an Ala side chain. The stippled surface indicates the space occupied by Val. Clearly, the tail of enacyloxin IIa is not hindered, but rather it makes a more extensive hydrophobic contact with the Val 375 side chain. In contrast, there is a steric clash for the kirromycin tail entering the hydrophobic pocket; the transparent van der Waals surface (purple) around the tail intersects with the Val 375(387) van der Waals envelope (glossed). C, selective differences in the tail moiety of the overlap between enacyloxin IIa (yellow) and kirromycin (pink), likely to affect the binding affinity and activity of the two antibiotics. The active core area of kirromycin is circled. the stability of the ␥-phosphate, which could be differently affected by the two antibiotics and which may also relate to the differential influence of cations on this activity.
In conclusion, the structural similarities and differences between enacyloxin IIa and kirromycin will not only explain central features of their action but also evoke selective approaches for obtaining novel inhibitors of protein synthesis, a proven target and a crucial aspect for society, taking into account the general increase in resistance toward antibiotics.