An Analysis of the Novel Fluorocycline TP-6076 Bound to Both the Ribosome and Multidrug Efflux Pump AdeJ from Acinetobacter baumannii

ABSTRACT Antibiotic resistance among bacterial pathogens continues to pose a serious global health threat. Multidrug-resistant (MDR) strains of the Gram-negative organism Acinetobacter baumannii utilize a number of resistance determinants to evade current antibiotics. One of the major resistance mechanisms employed by these pathogens is the use of multidrug efflux pumps. These pumps extrude xenobiotics directly out of bacterial cells, resulting in treatment failures when common antibiotics are administered. Here, the structure of the novel tetracycline antibiotic TP-6076, bound to both the Acinetobacter drug efflux pump AdeJ and the ribosome from Acinetobacter baumannii, using single-particle cryo-electron microscopy (cryo-EM), is elucidated. In this work, the structure of the AdeJ–TP-6076 complex is solved, and we show that AdeJ utilizes a network of hydrophobic interactions to recognize this fluorocycline. Concomitant with this, we elucidate three structures of TP-6076 bound to the A. baumannii ribosome and determine that its binding is stabilized largely by electrostatic interactions. We then compare the differences in binding modes between TP-6076 and the related tetracycline antibiotic eravacycline in both targets. These differences suggest that modifications to the tetracycline core may be able to alter AdeJ binding while maintaining interactions with the ribosome. Together, this work highlights how different mechanisms are used to stabilize the binding of tetracycline-based compounds to unique bacterial targets and provides guidance for the future clinical development of tetracycline antibiotics.

In order to better understand the determinants of how TP-6076 is recognized by AdeJ and the bacterial ribosome, cryo-electron microscopy (cryo-EM) was used to solve the structures of each complex. Here, we present the first structures of TP-6076 bound separately to the A. baumannii ribosome and the AdeJ pump. Additionally, we report the first structure of the AdeJ-TP-6076 complex to a resolution of 2.91 Å and three distinct structures of the ribosome-TP-6076 complex to resolutions of between 2.38 Å and 3.05 Å. Our results show that TP-6076 binds via very different modes of interaction with AdeJ and the ribosome. Compared to the tetracycline antibiotic eravacycline (Era)-bound structures, we show that modifications to the tetracycline core in positions 4 through 8 of TP-6076 can significantly alter binding to AdeJ with little to no change in its ribosome-bound position (27). Together, our work contributes key insights into the recognition of TP-6076 by AdeJ and the ribosome and provides a guide for future generations of tetracycline development with the goal of evading extrusion while maintaining antimicrobial function.

RESULTS
Structure of the A. baumannii AdeJ-TP-6076 complex. To determine how A. baumannii AdeJ recognizes the TP-6076 substrate, we utilized single-particle cryo-EM to determine its structure to high resolution. A. baumannii AdeJ was expressed in Escherichia coli, purified, and reconstituted into lipidic nanodiscs. The sample was incubated with TP-6076 to form the AdeJ-TP-6076 complex, and the structure of the complex was solved ( Fig. 1; see also Fig. S1 and Table S1 in the supplemental material). Overall, at least 1,046 residues of the 1,056 residues from full-length AdeJ were built into each model.
AdeJ assembles as a homotrimer and adopts the typical fold of a hydrophobeamphiphile efflux (HAE)-RND pump (26). In agreement with previous studies of AdeJ (27), each AdeJ protomer consists of a periplasmic and a transmembrane region. The transmembrane domain contains 12 transmembrane helices (TMs 1 to 12), while the periplasmic domain consists of a portal domain (PN1, PN2, PC1, and PC2) and a docking domain (DN1 and DN2) ( Fig. 1B to D).
Similar to our previous studies (27), a periplasmic cleft is formed between the PC1 and PC2 subdomains ( Fig. 1B and C). This cleft forms an entrance binding site for ligand interaction prior to shuttling past the flexible loop (F-loop) in the A. baumannii AdeB multidrug efflux pump (28). Presumably, the integrity of this site is important for drug recognition, where residues M666, L668, R701, R718, and T831 are coordinated to surround this entrance site. We emphasize that the corresponding residues of R718 in Escherichia coli AcrB (residue R717) and Pseudomonas aeruginosa MexB (residue R716) have previously been determined to be important for drug recognition (29,30).
The F-loop in RND-type efflux pumps leads the drug molecule from the entrance to the proximal binding site. As indicated previously (27), the F-loop in AdeJ consists of residues 671 to 680 that connect the cleft entrance to the proximal site. This loop is likely flexible, similar to the corresponding F-loop in the AdeB pump (28).
The gate loop (G-loop), which shuttles the ligand from the proximal to distal binding sites prior to extrusion, is formed by residues 615 to 624 in AdeJ. In AdeB, the G-loop has been shown to be flexible and occupy different conformations in apo and bound protomers (28). In addition, the conserved phenylalanine residue F612 of the AdeB G-loop (F618 in AdeJ) is actively engaged in anchoring the bound ligand at the distal site (28).
In the TP-6076-bound AdeJ trimer, each protomer occupies a distinct structural state forming an asymmetric trimer ( Fig. 1A to C). These different states are vital for drug export, as seen in previous fluorescence resonance energy transfer (FRET) studies of Campylobacter jejuni CmeB (37). Using three metrics for classification ( Fig. S2 and Table S2), the AdeJ-TP-6076 complex was determined to consist of binding, access, and extrusion protomers. This agrees with our previous studies of AdeJ bound to eravacycline and with previous cryo-EM studies of AdeB (27,28).
The entrance site of the access protomer of AdeJ is open, whereas its extrusion site is closed ( Fig. 1B and C). While a tunnel can be traced through the cleft entrance to the proximal binding pocket, a ligand is not found in the protomer.
Unlike the access protomer, the conformation of the extrusion protomer depicts that the periplasmic cleft between subdomains PC1 and PC2 is closed. Therefore, the entrance site of this protomer is also closed. An extrusion tunnel can be traced from the distal binding site out through the extrusion site, which is open in this conformational state. Presumably, this tunnel allows the ligand to be extruded from the distal binding pocket.
In the binding protomer, the entrance site between PC1 and PC2 is open ( Fig. 1B  and C). A binding tunnel can be traced from the entrance site through the F-loop, the proximal binding pocket, and the G-loop and into the distal binding pocket. This binding tunnel is the presumed path for the ligand to travel from the periplasmic entrance into the distal binding pocket. This agrees with previous molecular dynamics (MD) simulation studies of AdeB (28), where an ethidium molecule was found to follow this path upon binding to the distal pocket.
An extra density was observed in the binding protomer of AdeJ deep within the periplasmic cleft (Fig. 1D). This density corresponds to the bound TP-6076 antibiotic at the distal pocket of the binding protomer. The binding of TP-6076 is largely stabilized by both aromatic and hydrophobic interactions. Ten hydrophobic residues of the distal site anchor the four-ring system of the tetracycline core of the antibiotic ( Fig. 1D and E). These 11 residues are V139, F178, G179, F277, A326, Y327, F611, V613, F616, F618, and F629. Particularly, F178 and F616 participate in p -p stacking interactions with rings A and C, respectively. Additionally, V139 and F611 interact with the trifluoromethyl group of ring D, while G179, V613, and F277 participate in alkyl-alkyl and p-alkyl interactions with the diethylamino group on ring A. A326, Y327, and F629 are also engaged in alkyl-alkyl and p -alkyl interactions with the pyrrolidine group of ring D. Electrostatic interactions between Q87 and Q176 and the carboxamide group on ring A of the tetracycline core further stabilize TP-6076 binding to the distal pocket of AdeJ. As Mg 21 is known to coordinate with tetracyclines in the ribosome, it is somewhat surprising that no Mg 21 ions were found to bind the TP-6076 drug in the AdeJ-TP-6076 structure. However, this observation agrees with the X-ray structure of AcrB in complex with minocycline and the cryo-EM structure of AdeB in complex with eravacycline (27,38).
RND pumps utilize proton motive force (PMF) to extrude ligands from the periplasmic domain (27,28,35). In AdeJ, this force is most likely generated via the proton relay network established by the conserved residues D407, D408, K952, N953, and T989 in the transmembrane domain of each protomer. As in the previous studies of AdeB, AdeJ, and MtrD, our structural data indicate concerted motions between the rearrangement of hydrogen bonds in this proton relay network, where it couples with conformational changes between the resting, access, binding, and extrusion states of the AdeJ multidrug efflux pump (Fig. S2).
Structural analysis of the A. baumannii 70S ribosome. As tetracyclines target the bacterial 70S ribosome to inhibit protein synthesis, we solved the structure of TP-6076 bound to the A. baumannii ribosome to better understand the mode of the tetracycline-scaffold-ribosome interaction. The A. baumannii 70S ribosome was incubated with TP-6076 for 1 h, cryo-EM grids were prepared, and cryo-EM data were collected for this ribosome-TP-6076 complex. Using extensive classification, we were able to solve three separate cryo-EM structures of the A. baumannii 70S ribosome bound to TP-6076. These three 70S-TP-6076 structures are designated P-site tRNA, E-site tRNA, and empty 70S ribosome structures (Fig. S3).
The cryo-EM maps of these three ribosome structures were resolved to resolutions of between 2.38 Å and 3.05 Å ( Fig. S4 and Table S3), allowing the unambiguous determination of the 23S rRNA, 5S rRNA, and 28 rProteins of the 50S large ribosomal subunit and the 16S rRNA and 20 ribosomal proteins (20 rProteins) of the 30S small ribosomal subunit ( Fig. 2A to C). Using masked classification, a single tRNA molecule was found to occupy either the P site or the E site of the ribosome, giving rise to the P-site tRNA or E-site tRNA 70S cryo-EM structures, respectively. There was also a distinct class of single-particle population where no tRNA was found in the A, P, or E sites, leading to the empty 70S ribosome structure ( Fig. 2A to E).
Based on the structural information, there are two binding sites for TP-6076 present in all three 70S ribosome structures. These two TP-6076 binding sites are both located at the 30S small ribosomal subunit ( Fig. 3A to C). The first binding site depicts the putative tetracycline binding site, found in the 30S head ( Fig. 3A to C). At this site, one TP-6076 molecule occupies this putative binding site stabilized by coordination with two Mg 21 ions and a number of electrostatic interactions from nucleotides rU962, r2MG963, rG1050, rC1051, rC1192, rA1193, rA1194, and rG1195. This site overlaps the A-site tRNA pocket of the 30S subunit, posing the possibility that the drug is responsible for blocking the attachment of tRNA to the A site and halting translation elongation, a trademark characteristic of tetracycline antibiotics (6).
The second TP-6076 binding site is located within the core of the small 30S ribosomal subunit (Fig. 3A to C). At this position, two TP-6076 molecules coordinate with two Mg 21 ions in the drug binding site. These drug molecules are stabilized by a number of electrostatic interactions via nucleotides rA569, rA570, rA812, rA912, rU913, rU1390, rA1391 rA1496, rA1497, rG1505, rC1506, rG1526, and rG1527. As with the first drug binding site, this site remains consistent throughout the three 70S structures. These two TP-6076 molecules directly interact with helix 44 (h44) of the 16S rRNA, which is vital for translational fidelity and initiation (39). While mutation studies have confirmed the putative tetracycline binding site within the 30S head (40), no systemic studies have been performed to show the importance of the second TP-6076 binding site within the core of the 30S subunit. Whether binding at this site by TP-6076 has biological implications invites further experimentation.

DISCUSSION
Antibiotic resistance is a major clinical challenge for modern medicine. A. baumannii is a major threat in this regard, where carbapenem resistance is now present in most A. baumannii strains along with an alarming rise in tigecycline and colistin resistance (41,42). One major reason for this is the overexpression and selective mutation of multidrug efflux pumps (26,43). Therefore, a greater understanding of this resistance mechanism will aid in the future development of inhibitors that effectively block this efflux process and increase the potency of current antibiotics.
Toward this goal, we have solved the structures of TP-6076 bound to both the A. baumannii multidrug efflux pump AdeJ and the A. baumannii ribosome using cryo-EM. This adds to our recent work in which the structures of AdeJ and the A. baumannii ribosome bound to the tetracycline antibiotic eravacycline (Era) were solved (27,44,45). Together, these data highlight the versatility of the multidrug efflux pump AdeJ and its ability to recognize a wide variety of substrates.
TP-6076 binding to AdeJ is accomplished mainly through hydrophobic interactions, with additional stabilization via two unique electrostatic interactions. To accomplish TP-6076 Binding to the A. baumannii Ribosome and AdeJ ® this, the hydrophobic face of the TP-6076 molecule containing positions 4 to 7 is utilized ( Fig. 4A and C). All hydrophobic residues within the distal binding pocket of AdeJ (V139, F178, G179, F277, A326, Y327, F611, V613, F616, and F629) participate in both Era and TP-6076 binding. Surprisingly, in order to better accommodate the aminoethyl and trifluoromethyl ring modifications and the shorter pyrrolidine group, TP-6076 is flipped relative to Era in AdeJ (Fig. 4A, C, E, and G). This flipped pose then allows Q87 and Q176 to form additional electrostatic interactions with the carboxamide group at position 2 of ring A. This flipped pose, while still utilizing the same hydrophobic face of the tetracycline moiety, highlights the versatility of the AdeJ pump in substrate recognition and extrusion.
In the ribosome, TP-6076 occupies a site nearly identical to that of Era in the putative tetracycline binding site in the head of the 30S ribosomal subunit (Fig. 3A to C and Fig. 4B, D, F, and H). In both structures, two Mg 21 ions coordinate with the tetracycline core, and binding is stabilized through favorable electrostatic interactions with residues r2MG963, rG1050, rC1051, rC1192, rA1193, rA1194, and rG1195. This shows that despite modifications to the drug, TP-6076 likely maintains tetracycline antimicrobial properties, in agreement with previous biological studies (4,5).
Unlike Era, where a second ligand binding site is found in the 50S large ribosomal subunit, a second TP-6076 site is found in the core of the 30S small ribosomal subunit. In this site, two TP-6076 molecules coordinate with two Mg 21 ions and are stabilized by a number of electrostatic interactions. The TP-6076 residues at this site interact directly with h44 of the 16S rRNA. This helix is responsible for maintaining coding fidelity and is the main target of aminoglycoside antibiotics such as kanamycin and gentamicin (10,11,39). Therefore, this site may play a role in the overall efficacy of TP-6076 and invites further experimentation.
It is known that the binding affinities of drugs for multidrug efflux pumps are in the micromolar range (28,34,37,(46)(47)(48)(49)(50). Although the strengths of antibiotic binding within the 70S ribosomes are not as commonly studied, there is evidence that these binding affinities can be substantially strong in comparison. For example, the ribosomes bind macrolides with dissociation constants in the nanomolar range (51,52). However, the ribosomes also bind several drugs, including tetracyclines (53) and aminoglycosides (54), within the micromolar range, where these binding affinities are similar to those for multidrug efflux pumps.
From our cryo-EM data, it is clear that AdeJ utilizes the hydrophobic properties of tetracyclines for recognition, while its electrostatic properties are vital for interactions with the ribosome. This conserved mode of recognition is evident when comparing the binding of both TP-6076 and Era to AdeJ (Fig. 4). Interestingly, in order to accommodate modifications to the hydrophobic face of TP-6076, the ligand is present in a flipped orientation relative to Era (Fig. 4A, C, E, and G) in AdeJ, while the ribosome binding site is nearly identical (Fig. 4B, D, F, and H). As the tetracycline class of antibiotics utilizes electrostatic interactions to bind and occupy the same respective putative tetracycline site in the 70S ribosomes (11), and both TP-6076 and Era employ hydrophobic interactions to specifically contact the AdeJ pump, we believe that this trend would remain consistent with most tetracycline drugs. Our experimental result underscores the phenomenon that the predominant interaction between tetracyclines and 70S ribosomes is electrostatic in nature, whereas the prevalent association between tetracyclines and multidrug efflux pumps is hydrophobic in character. This notable difference in substrate recognition between 70S and AdeJ may be exploited in future tetracycline-based drugs, where modifications to only the hydrophobic side of the tetracycline core will significantly hinder recognition by multidrug efflux pumps like AdeJ but maintain binding to the ribosome and, therefore, its antimicrobial properties.

MATERIALS AND METHODS
Expression and purification of A. baumannii AdeJ. AdeJ from A. baumannii AB0057 was expressed and purified as previously described (27). In short, AdeJ was cloned into a pET15b vector with a 6ÂHis tag at the N terminus and confirmed by DNA sequencing. The plasmid was transfected into E. coli BL21 (DE3) DacrB cells and grown in 6 L of lysogeny broth (LB) medium with 100 mg/mL of ampicillin at 37°C. At an optical density at 600 nm (OD 600 ) of 0.5, the expression of AdeJ was induced with 0.2 mM isopropyl-b-D-thiogalactopyranoside (IPTG). Bacteria were harvested after 4 h of induction. Cells were resuspended in low-salt buffer (100 mM sodium phosphate [pH 7.4], 10% glycerol, 5 mM EDTA, and 1 mM phenylmethanesulfonyl fluoride [PMSF]) and disrupted with a French pressure cell. The collected membrane fraction was washed twice with high-salt buffer (20 mM sodium phosphate [pH 7.4], 2 M KCl, 10% glycerol, 5 mM EDTA, and 1 mM PMSF) and once with final buffer (20 mM Na-HEPES [pH 7.5] and 1 mM PMSF). Membrane protein was solubilized using 2% (wt/vol) n-dodecyl-b-D-maltoside (DDM) for 3 h. Insoluble material was removed by centrifugation at 100,000 Â g, and the extracted protein was purified using a Ni-nitrilotriacetic acid (NTA) column. The purity of AdeJ (.95%) was determined via SDS-PAGE TP-6076 Binding to the A. baumannii Ribosome and AdeJ ® gels stained with Coomassie brilliant blue. The purified protein was then dialyzed against 20 mM Na-HEPES (pH 7.5) and concentrated to 7 mg/mL (60 mM) in a solution containing 20 mM Na-HEPES (pH 7.5) and 0.05% DDM.
AdeJ-nanodisc preparation. AdeJ was assembled into nanodiscs using a mixture of 20 mM AdeJ, 45 mM membrane scaffold protein (MSP) (1E3D1), and 930 mM E. coli total lipid extract. The mixture was incubated at room temperature for 15 min. To remove the detergent, 0.8 mg/mL prewashed Bio-Beads (Bio-Rad) was added. The mixture was incubated on ice for 1 h, followed by incubation at 4°C overnight. The sample was filtered using a 0.22-mm nitrocellulose filter to remove Bio-Beads and purified from empty nanodiscs using a Superose 6 column (GE Healthcare) in a solution containing 20 mM Tris-HCl (pH 7.5) and 100 mM NaCl. Fractions containing the trimeric AdeJ-nanodisc complex were collected for cryo-EM sample preparation.
Purification of the A. baumannii ribosome. Ribosomes were purified directly from A. baumannii AB0057 as previously described (55). In brief, A. baumannii cells were lysed in a solution containing 20 mM Tris (pH 7.5), 50 mM magnesium acetate (MgOAc), 100 mM NH 4 Cl, 1 mM dithiothreitol (DTT), and 0.5 mM EDTA (56) using a French pressure cell. The lysate was centrifuged at 30,000 Â g for 30 min to pellet insoluble material. Ribosomes were pelleted from the supernatant using a sucrose cushion buffer (lysis buffer with 1.1 M sucrose) and centrifuged for 16 h at 100,000 Â g. Ribosome-containing pellets were stored at 280°C until further use.
Sucrose gradient purification using a sucrose gradient buffer (20 mM HEPES, 14 mM MgOAc, 100 mM KCl, 0.2 mM DTT, and 0.1 mM PMSF with 10% and 40% sucrose and mixed using a gradient maker) and 16 h of centrifugation at 100,000 Â g was used to increase 70S purity. Fractions containing 70S particles were collected, concentrated, exchanged into ribosome buffer (20 mM HEPES-KOH [pH 7.5], 10 mM MgOAc, and 100 mM KCl), aliquoted, flash-frozen in liquid nitrogen, and stored at 280°C until further use.
Cryo-EM sample preparation. A total of 10 mM AdeJ-nanodisc was incubated with 20 mM TP-6076 for 1 h prior to grid preparation to create the AdeJ-TP-6076 complex. The sample was applied to glowdischarged holey carbon grids (Quantifoil Cu R1.2/1.3, 300 mesh), blotted for 5 s, and then plunge-frozen in liquid ethane using a Vitrobot (Thermo Fisher). For the A. baumannii ribosome, 100 nM ribosome sample was incubated with 50 mM TP-6076 for 2 h prior to grid preparation to create the ribosome-TP-6076 complex. Grids were plunge-frozen with a Vitrobot using a 15-s blot time and graphene oxide-coated Quantifoil R1.2/1.3 grids. All grids were transferred to cartridges prior to data collection.
Data collection. For the AdeJ-TP-6076 complex, data were collected on a Titan Krios system equipped with a K3 direct electron detector (Gatan). Data were collected in superresolution mode at a magnification of Â81,000, resulting in a physical pixel size of 1.08 Å/pix (superresolution of 0.54 Å/pix). Each micrograph was collected over 40 frames with a total dose of 36 e 2 /Å 2 over 2 s using SerialEM (57). For the ribosome-TP-6076 complex, data were collected on a Titan Krios system equipped with a Falcon III detector (Gatan). Data were collected in linear mode at a magnification of Â75,000, corresponding to a physical pixel size of 1.089 Å/pix. Each micrograph was collected over 60 frames over 1.9 s with a total dose of 50 e 2 /Å 2 using EPU (Thermo Fisher Scientific).
Data processing. For AdeJ, superresolution stacks were aligned and binned by 2 using patch motion, and contrast transfer function (CTF) was estimated using patch CTF in cryoSPARC v3 (58). Blob picker followed by two-dimensional (2D) classification was used to generate templates for template picker (see Fig. S1 in the supplemental material), which picked an initial pool of 2,975,155 particles. 2D classifications followed by ab initio and heterogeneous three-dimensional (3D) classifications were used to remove false picks and contamination from ice and carbon. 3D classification using 3D variability analysis on the remaining 595,961 particles resulted in a single high-resolution class for refinement. Nonuniform refinement followed by masked local refinement with nonuniform sampling resulted in 2.91-Å-resolution maps of the AdeJ-TP-6076 complex based on the gold-standard Fourier shell correlation (GS-FSC) ( Fig. S1 and Table S1). The final maps were modified using resolve cryo-EM (59) from the phenix suite of programs (60).
For the ribosome, image stacks were motion corrected using patch motion, and patchCTF was used to estimate CTF (58). Blob picker using a subset of micrographs was used to generate templates for template picker, resulting in an initial particle stack of 725,635 particles (Fig. S3). 2D classifications followed by ab initio and heterogeneous 3D classifications were used to generate initial templates. Using a modified "build-and-retrieve" approach (61), initial templates were used to retrieve particles from the initial particle stack, resulting in markedly higher final particle counts. 70S classes were separated based on tRNA populations in the P and E sites using 3D variability analysis in cryoSPARC (58) and refined using homogeneous refinement. The final refinement of each 70S class was performed by using masked local refinement with nonuniform sampling and dividing each map into three sections: 50S, 30S core, and 30S head (Fig. S4). Final maps were sharpened using resolve cryo-EM in phenix (59). Locally refined map segments were used to build the final model for each class (70S P-site tRNA, 70S E-site tRNA, and empty 70S), and the final composite map for visualization was created using vop maximum in chimera (62).
Model building and refinement. Models for the TP-6076-AdeJ complex and the ribosome (70S P-site tRNA, 70S E-site tRNA, and empty 70S) were based on the cryo-EM maps. Previously determined structures for AdeJ (27) and the A. baumannii ribosome (55) were used as starting models. Starting models were aligned to the cryo-EM maps using chimera (62). The width of the extrusion tunnel along with the periplasmic cleft state and PMF measurements were used to classify protomer states ( Fig. S2 and Table S2). For the ribosome, Mg 21 ions directly coordinated with each TP-6076 molecule were manually added in Coot (63). Models for all structures were built using Coot (63) and refined using phenix.real_space_refine (64) from the phenix suite (60). Final structures were evaluated using MolProbity (65), with final refinement statistics included in Tables S1 and S3.
Data availability. Atomic coordinates and the EM map for the AdeJ-TP-6076 complex have been deposited in the RCSB Protein Data Bank (PDB) under accession number 7RY3 and the Electron Microscopy Data Bank (EMDB) under accession number EMD-24732. Atomic coordinates for TP-6076 in complex with A. baumannii P-site tRNA 70S, E-site tRNA 70S, and empty 70S have been deposited in the PDB under accession numbers 7RYF, 7RYG, and 7RYH and in the EMDB under accession numbers EMD -24738, EMD-24739, and EMD-24740, respectively.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.