Antibiotic resistance ABCF proteins reset the peptidyl transferase centre of the ribosome to counter translational arrest

Abstract Several ATPases in the ATP-binding cassette F (ABCF) family confer resistance to macrolides, lincosamides and streptogramins (MLS) antibiotics. MLS are structurally distinct classes, but inhibit a common target: the peptidyl transferase (PTC) active site of the ribosome. Antibiotic resistance (ARE) ABCFs have recently been shown to operate through direct ribosomal protection, but the mechanistic details of this resistance mechanism are lacking. Using a reconstituted translational system, we dissect the molecular mechanism of Staphylococcus haemolyticus VgaALC and Enterococcus faecalis LsaA on the ribosome. We demonstrate that VgaALC is an NTPase that operates as a molecular machine strictly requiring NTP hydrolysis (not just NTP binding) for antibiotic protection. Moreover, when bound to the ribosome in the NTP-bound form, hydrolytically inactive EQ2 ABCF ARE mutants inhibit peptidyl transferase activity, suggesting a direct interaction between the ABCF ARE and the PTC. The likely structural candidate responsible for antibiotic displacement by wild type ABCF AREs, and PTC inhibition by the EQ2 mutant, is the extended inter-ABC domain linker region. Deletion of the linker region renders wild type VgaALC inactive in antibiotic protection and the EQ2 mutant inactive in PTC inhibition.


INTRODUCTION
Approximately half of the antibiotics currently in use for treating bacterial infections inhibit protein synthesis, predominantly by targeting key functional sites of the ribosome (1). To counteract antibiotics, bacteria have developed an array of resistance mechanisms (2). These mechanisms can be broadly classified into two categories. The first strategy is to decrease the intracellular concentration of the active antibiotic. This can be achieved by preventing antibiotic uptake (3), by actively excreting the antibiotic (4) or by inactivating the drug via chemical modification or degradation (5). The second strategy is to render the pathway that is targeted by the antibiotic immune to the drug. This can be achieved by mutation or modification of the antibiotic target (6). Alternatively, a resistance factor can directly interact with the antibiotic target to displace the antibiotic or prevent its binding in the first place, as exemplified by the ribosome protection factors TetO and TetM (7,8). These members of the EF2 family of translational GTPases share the domain structure of elongation factor EF-G (9). Like EF-G, Tet proteins bind to the A-site of the ribosome, where they sterically clash with tetracycline bound to the ribosome, thus effectuating antibiotic dissociation from its target (10,11).
Antibiotic resistance proteins in the ABC (ATP-binding cassette) superfamily of ATPases are well known for conferring resistance through the first of the two strategies, i.e. by reducing the concentration of the drug in the cell by pumping it out (4). However, the crucial component of the classical ABC pump, its transmembrane domain (12,13) is lacking in antibiotic resistance factors that belong to the ABCF subfamily of ABC ATPases (14) (BioRxiv: https://www.biorxiv.org/content/early/2017/ 11/16/220046). This raises the question of whether resistance is achieved instead via the second strategy described above i.e. via direct protection of the inhibited pathway.
Two archetypal antibiotic resistance (ARE) ABCF factors are VgaA and LsaA, which carry two nucleotide binding domains, separated by a linker region (Figure 1). Staphylococcal VgaA (standing for 'virginiamycin A-like antibiotic resistance') is a plasmid-encoded resistance factor described in Staphylococcus aureus (15), Staphylococcus epidermidis (16) and Staphylococcus haemolyticus (17 sistance') is encoded on the chromosome of Enterococcus faecalis (18,19). The two terms, 'virginiamycin A-like antibiotics' and 'streptogramin A' are synonymous and refer to the same chemical group of compounds (20). VgaA and LsaA confer protection against the same antibiotic classes, streptogramin A and lincosamides (so-called LS A phenotype) and pleuromutilins, such as tiamulin, although the relative activity against different antibiotics varies (15)(16)(17)(18)(19)21,22). An extreme example is S. haemolyticus VgaA LC , which has a substrate specificity that is strongly shifted towards lincosamides, such as lincomycin and its derivative, clindamycin (17). While chemically unrelated, pleuromutilin, streptogramin A and lincosamide antibiotics bind to a common target, the peptidyl transferase centre (PTC) of the large (50S) ribosomal subunit (23)(24)(25)(26)(27)(28)(29). The PTC is the catalytic centre of the ribosome, where during each iteration of the elongation cycle the amino acid moiety of an incoming aminoacyl-tRNA to the ribosomal A-site forms a peptide bond with the peptidyl-tRNA in the P-site (30). As the ARE ABCFs confer resistance to antibiotics that are chemically unrelated but share a binding target, it was hypothesized that AREs directly interact with the ribosome to displace the drug (14). A decade later this hypothesis has been experimentally substantiated both for S. aureus VgaA and E. faecalis LsaA (31). While the molecular mechanism of resistance is still unknown, an inter-ATPase domain linker region that protrudes towards the PTC in the ribosome-binding translation factor ABCF EttA (32,33) is extended in AREs (Supplementary Figure S1). Therefore, it was hypothesized that the linker could directly displace the antibiotic from the PTC (21,34). Indeed, mutations in this region affect the antibiotic specificity of the ARE (21). These initial discoveries have set the stage for dissecting the system biochemically using a reconstituted translational system.
Here, we use antibiotic inhibition of transpeptidation of S. aureus 70S ribosomal initiation complexes and rescue by S. haemolyticus VgaA LC to probe three aspects of the molecular mechanism of ARE ABCFs. First, we addressed the question of nucleotide substrate specificity. Not all the ABC enzymes are strict ATPases, and certain representatives are, essentially, NTPases that in addition to ATP can utilize GTP, CTP and UTP (35). Second, we clarified the functional role of NTP hydrolysis in the functional cycle of VgaA LC . NTPases can operate either as a 'molecular switch' where NTP versus NDP binding alters the structural conformation of the enzyme thereby exerting its function, or as a 'molecular machine', which uses NTP to NDP hydrolysis to drive a 'powerstroke' fueling the performed mechanical work. Both versions are present in the ABC family (12). Finally, we probed structure-function relationships, focusing on the extended alpha-helical linker region connecting the two ATP cassette domains of antibiotic resistance ABCFs ( Figure 1).

Phylogenetic analysis
Sequences of previously documented ABCF AREs were retrieved from UniProt (36) and the Comprehensive Antibiotic Resistance Database (CARD) (37). Sequences were aligned with Mafft v7.164b with the l-ins-i and maximum likelihood phylogenetic analysis was carried out with RAxML-HPC v.8 (38) on the CIPRES Science Gateway v3 (39) with 100 bootstrap replicates and the LG model of substitution. Alignment positions with >50% gaps were excluded from the analysis.

Protein cloning, expression and purification
All cloning was performed by the Protein Expertise Platform at Umeå University. S. haemolyticus VgaA LC ORF was PCR-amplified from pRB374 VgaA LC plasmid (17) and sub-cloned to the pET24d expression vector with a Cterminal 6His tag preceded by a single glycine linker. E. faecalis LsaA ORF was PCR amplified from pTEX5333 plasmid (18) and sub-cloned into pCA528 vector for 6His-SUMO-tagging (40). ATPase deficient (EQ 2 ) mutants were generated by introducing E105Q and E410Q (VgaA LC ) or E142Q and E452Q (LsaA) point mutations. The linker deletion ( L) VgaA LC mutants were generated by preplacing K199-A226 with GSG in ether wt or EQ 2 VgaA LC .
The purity of protein preparations was assessed by SDS-PAGE and spectrophotometrically (OD 280 /OD 260 ratio of ≈1.8 for VgaA LC and 1.6-1.8 for LsaA). The proteins were aliquoted and stored at -20 • C.

Puromycin reaction
The puromycin reaction was carried out at 37 • C in HEPES:Polymix pH 7.5 2+ buffer (41). The biochemical system from purified E. coli components has been described earlier (42,43). 35 S-Met-puromycin release was followed using 10% TCA precipitation and centrifugation with scintillation counting of the supernatant ( 35 S-Met-puromycin) and pellet (intact 35 S-fMet-tRNA i fMet ). The percentage of 35 S-fMethionine released from the 70S IC was calculated by dividing the signal from the supernatant by the sum of signals from supernatant and pellet. The data are presented as geometric means, and error bars represent the range of experimental values.
Detailed experimental protocols, buffer preparations, expression and preparation of initiation factors and Ulp1 protease are found in SI Materials and Methods.

S. haemolyticus VgaA LC ATP-dependently rescues the puromycin reactivity of S. aureus initiation complexes inhibited by lincomycin, clindamycin, virginiamycin M1 or tiamulin
We used the so-called puromycin reaction to assay transpeptidation functionality of the ribosome. The aminonucleoside antibiotic puromycin is a structural mimic of the 3 end of aminoacyl-tRNA that prematurely terminates translation by substituting for the incoming aminoacyl-tRNA and acting as an acceptor either for the growing peptide chain donated by the P-site peptidyl-tRNA, or, during initiation, the fMet donated by P-site initiator tRNA (fMet-tRNA i fMet ) (Figure 2A) (44). Using the puromycin reactivity of 35 S-methionine-labeled 70S initiation complexes (70S IC) as a biochemical system, we followed the inhibition of PTC reactivity by antibiotics and PTC rescue by ABCF ARE factors. The 70S IC was formed from purified 70S ribosomes isolated from either E. coli, S. aureus or E. faecalis, a synthetic mRNA encoding the Met-Phe dipeptide (mRNA(MF)), and E. coli initiator tRNA that is aminoacylated with 35 S-methionine and subsequently formylated ( 35 S-fMet-tRNA i fMet ) (42). Formation of the 70S ICs was catalyzed by E. coli initiation factors that were subsequently removed by ultracentrifugation through the sucrose cushion.
Since E. coli 70S ribosomes are intrinsically resistant to lincomycin (27,45), we opted for S. aureus 70S purified from the SH1000 model laboratory strain (46). Unlike E. coli, S. aureus ribosomes are highly sensitive to lincosamide antibiotics which are used to treat Staphylococcal infections in clinical practice (47), and several AREs have been identified as resistance factors in these bacteria. Already in moderate excess over ribosomes  In the absence of ATP (empty red circles) addition of VgaA LC has no effect. All experiments were performed at 37 • C in HEPES:Polymix pH 7.5 buffer, 5 mM Mg 2+ . 0.5 M S. aureus IC(MF) were programmed with synthetic mRNA(MF) encoding Met-Phe dipeptide and E. coli 35 S-fMet-tRNA i fMet (occupancy 60-80%). ICs were incubated with lincomycin for 5 min to allow equilibrium binding to be achieved prior the addition of puromycin. The data are presented as geometric means, and error bars represent the range of experimental values. 35 S-methionine is released after 1000 s, reflecting slow dissociation of the drug followed by the puromycin reaction. Clindamycin near-completely inhibits puromycin reactivity even after a 1000-s incubation with puromycin, indicating even slower dissociation. Conversely, in the case of intrinsically lincomycin resistant E. coli 70S IC, puromycin releases 35 S-methionine even in the presence of 5 or 15 M lincomycin, resulting in a near-complete deacylation of the P-site 35 S-fMet-tRNA i fMet already by 200 seconds (Supplementary Figure S2A).
Since puromycin reacts with the P-site substrate relatively slowly, with a maximum rate of a 50 s −1 at pH 7.5 (48), it could potentially allow rebinding of the PTC blocker and, therefore, fail to reliably probe the dissociation kinetics of the antibiotic. Therefore, we tested the kinetic competitiveness of 1 mM puromycin against lincomycin added at the saturating concentration, 1 M ( Supplementary Figure S2B). When lincomycin is preincubated with ribosomes prior to the addition of puromycin, the puromycin reactivity is completely abrogated. However, when lincomycin is added simultaneously with puromycin, the PTC blocker abrogates a mere 5% of puromycin reactivity. This shows that when added in great excess over lincomycin, puromycin is a considerably faster binder and is, indeed, a reliable tool for probing the antibiotic dissociation from the ribosome.
Having established that the system is reliable, we tested the effect of 1 M S. haemolyticus VgaA LC on puromycin reactivity of S. aureus 70S IC inhibited by saturating concentrations of lincomycin ( Figure 2D) and clindamycin ( Figure 2E). In the absence of nucleotides, VgaA LC has no effect on the puromycin kinetics. Conversely, in the presence of 1 mM ATP, VgaA LC rescues the puromycin reactivity, although not completely. Bacterial antibiotic susceptibly tests demonstrated that S. haemolyticus VgaA LC has specificity strongly shifted towards lincosamides and displays relatively poor activity against streptogramin A antibiotics (17). We have tested in our biochemical system a representative of the streptogramin A class, virgniamycin M1. The antibiotic is a poorer inhibitor of the puromycin reaction, requiring 10 M concentration for complete inhibition at the 10 s time point (Supplementary Figure S3A). While addition of VgaA LC has a clear ATP-dependent protective effect (Supplementary Figure S3C), the effect is not as pronounced as in the case of lincomycin. The pleuromutilin tiamulin is a moderately more efficient PTC inhibitor than virgniamycin M1, with inhibition efficiency saturating at 10 M (Supplementary Figure S3B). Again, VgaA LCmediated protection is strictly dependent on the presence of ATP (Supplementary Figure S3D).

VgaA LC is an NTPase that operates as a molecular machine, not as a molecular switch
In the following experiments we focused on ABCFmediated protection from lincomycin. The addition of 1 M VgaA LC and 1 mM ATP rescues the puromycin reactivity inhibited by saturating concentration of the antibiotic (1 M). However, the protection is not complete; the transpeptidation reaction is still significantly slower than in the absence of the antibiotic. One possible reason could be that the system is not saturated with either the nucleotide substrate, ATP, or VgaA LC . Therefore, we titrated ATP ( Figure 3A) and VgaA LC ( Figure 3B) in the presence of 1 M lincomycin. The protective effect reaches a plateau at 0.5 mM ATP and 0.5 M VgaA LC , i.e. 1:1 ratio of ARE to 70 IC. Even in these conditions, puromycin reactivity is not completely restored. In the following puromycin reaction assays we used excess concentrations of VgaA LC and nucleotides: 1 M and 1 mM, respectively.
Next, we characterized the nucleoside triphosphate specificity of VgaA LC -mediated lincomycin resistance. The puromycin release kinetics is identical when ATP is substituted for GTP, CTP or UTP, suggesting that VgaA LC is an NTPase (Figure 3C). In the absence of VgaA LC none of the NTPs alone have an effect on the system (Supplementary Figure S4A). To characterize the NTP specificity of VgaA LC , we titrated GTP, the second most-abundant NTP species in bacteria after ATP (49) (Supplementary Figure S4B). Similarly to ATP, the full activity of VgaA LC is achieved at 0.5 mM GTP, suggesting that in the cell the enzyme operates at saturating concentrations of both NTP species. Finally, to probe the role of NTP hydrolysis by VgaA LC , we compared ATP, its non-hydrolysable analogue ADPNP (adenosine 5 -(␤-␥ -imido) triphosphate, App(NH)p), a slow-hydrolysable analogue ATP␥ S (adenosine 5-O-(3-thio) triphosphate) and the product of ATP hydrolysis, ADP ( Figure 3D). Neither ADPNP, nor ATP␥ S, nor ADP support VgaA LC activity, suggesting driving the conformational switch by ATP analogues is not sufficient to rescue the PTC activity.
Since VgaA LC can utilize GTP as a substrate, we tested if the alarmone nucleotide ppGpp could suppress the VgaA LC -mediated resistance acting as an orthosteric inhibitor the same way as it inhibits GTPases and nucleotide metabolism enzymes (50). During acute amino acid starvation, ppGpp becomes the dominant guanosine nucleotide, reaching sub-mM concentration (49). However, even when added at 0.5 mM, ppGpp does not have a pronounced inhibitory effect on VgaA LC (Supplementary Figure S5).

The hydrolytically inactive S. haemolyticus VgaA LC EQ 2 and E. faecalis LsaAEQ 2 mutants inhibit ribosomal peptidyl transferase in the presence of NTP
Use of non-or slowly-hydrolysable ATP analogues is one approach to generate an ATP-bound state of VgaA LC . An alternative is to mutate conserved catalytic glutamate residues following the Walker B motif in the two ABC cassettes and use the native ATP substrate. These glutamates form strong hydrogen bond interactions with the attacking water molecule, polarizing it as a nucleophile (51). Simultaneous mutation of the residues in both cassettes for glutamine (E105Q and E410Q; EQ 2 ) results in a deficiency in ATP hydrolysis by ABC ATPases (52) and locks the enzyme in an ATP-bound active conformation (33,53).
The hydrolytically deficient VgaA LC EQ 2 :ATP does not rescue inhibition of transpeptidation by lincomycin and compromises protection by the wild type VgaA LC (Figure 4A). A possible explanation is that NTPase-inactive VgaA LC EQ 2 competes with the wild type for binding to the ribosome, thus preventing the wild type from carrying out its protective function. A more non-trivial scenario is that The resistance effect saturates at 0.5 mM ATP and 0.5 M VgaA LC, respectively. VgaA LC -mediated rescue of puromycin reactivity is supported by either ATP, GTP, CTP or UTP nucleotides added at 1 mM, suggesting that VgaA LC is an NTPase, rather than a strict ATPase (C). Neither ADP, nor the nonhydrolysable ATP analogue ADPNP nor the slow-hydrolysable analogue ATP␥ S can support VgaA LC activity, indicating that NTP hydrolysis, not just the NTP binding, are necessary (D).
in addition to competing with the wild type, VgaA LC EQ 2 actively inhibits transpeptidation. Therefore, we tested the effects of VgaA LC EQ 2 on transpeptidation in the absence of lincomycin. The addition of increasing concentrations of VgaA LC EQ 2 in the presence of ATP progressively inhibits the puromycin reaction, reducing it by ≈70% in the presence of 2 M protein ( Figure 4B). Just like the protective effect of the wild type protein, PTC inhibition by VgaA LC EQ 2 requires NTP (ATP and GTP were tested), and is not supported by ADP, ADPNP or ATP␥ S ( Figure 4C).
To test whether PTC inhibition is a general feature of EQ 2 versions of ABCF ARE factors, we characterized E. faecalis LsaA. Although the efficiency is lower than S. haemolyticus VgaA LC , wild type LsaA protects E. faecalis 70S IC from inhibition by lincomycin ( Figure 5A). Similarly to VgaA LC , The EQ 2 mutant inhibits the transpeptidation activity of E. faecalis 70S IC in the presence of ATP ( Figure 5B). Note that unlike E. faecalis LsaA or S. haemolyticus VgaA LC , the EQ 2 mutant of E. coli housekeeping ABCF EttA does not inhibit the PTC activity directly as manifested by unperturbed formation of the first peptide bond (32).

Deletion of the linker region renders wild type VgaA LC inactive in antibiotic protection and the EQ 2 mutant inactive in PTC inhibition
The likely structural candidate responsible for antibiotic displacement by wild type VgaA LC and PTC inhibition by the EQ 2 mutant is the extended inter-ABC linker region (21) (Figure 1). This hypothesis, however, has never been tested experimentally. In the absence of structural information on VgaA LC it is impossible to design a 'clean' deletion of the linker region. We designed a deletion that reduces the length of the linker to that seen for EttA. In this truncated VgaA LC L mutant, amino acids K199-S226 are substituted for a 'stump' formed by a flexible GSG linker. Additionally, the L deletion was combined with the EQ 2 double point mutation (E105Q and E410Q) to inactivate hydrolytic activity. Deletion of the linker region has a profound effect on the functionality of VgaA LC : neither does VgaA LC L protect from lincomycin nor does VgaA LC EQ 2 L compromise the protective effect of the wild type VgaA LC ( Figure 4D). This suggests that the truncated protein can no longer reach and interact with the PTC region (Supplementary Figure S1). However, it is also possible that the deletion alters the overall structural integrity of VgaA LC thus inhibiting the protein's activity indirectly,  Figure 4. The linker region is essential for NTP-dependent inhibition of ribosomal peptidyl transferase activity by the NTP hydrolysis-incompetent VgaA LC EQ 2 mutant. The NTPase-incompetent VgaA LC EQ 2 mutant does not rescue the puromycin reactivity of S. aureus 70S IC(MF) inhibited by 1 M lincomycin (dark blue filled circles), and compromises protection by the wild type protein (dark blue circles, red fill) (A). The VgaA LC EQ 2 mutant inhibits IC(MF) puromycin reactivity in the absence of antibiotics in the presence of 1 mM ATP (B and C) or GTP (C), but not 1 mM ADP, ATP␥ S or ADPNP (C). The VgaA LC L mutant in which inter-ABC linker region (amino acids K199-S226) is substituted for a GSG linker neither protects the puromycin reactivity from 1 M lincomycin (filled green circles), and its EQ 2 variant, VgaA LC EQ 2 L, does not abolish the protective effect of the wild type protein (black circles with red fill) (D). e.g. by abrogating its binging to the ribosome or/and AT-Pase activity. Detailed structural information is necessary to guide precise structure-function characterization of ARE ABCF factors.

DISCUSSION
We suggest a model of ARE-mediated ribosome protection from PTC inhibition by LS A antibiotics ( Figure 6). Powered by NTP-hydrolysis, ARE factors dislodge LS A antibiotics from the ribosome. In the cell, AREs can use any NTP species, the most abundant being GTP and ATP. The association of the NTP-bound ARE with the ribosome causes transient arrest of transpeptidation, conceivably via a direct contact of the linker region with the PTC.
The relaxed nucleotide substrate specificity seen with VgaA LC is not unusual for ABC enzymes, e.g. E. coli transporter CvaB uses both GTP and ATP (54), and ribosomeassociated ABCF EttA was mentioned by Böel and colleagues to interact with both ATP and GTP (no data shown in the original publication) (32). To achieve its maximum activity, VgaA LC requires ≥500 M of the NTP substrate. While this is a relatively high concentration, it follows the general trend of enzyme affinity to substrate correlating with the concentration of substrate (55): there is no evolutionary pressure selecting for high affinity to abundant substrates. Since two NTP species--GTP and ATP--are present in bacterial cell in the mM concentration range (49), it is not surprising that in the test tube VgaA LC requires high NTP concentration.
Our biochemical analyses raise several pertinent questions that can be addressed by structural and microbiological approaches. First, high resolution structural information is necessary to unequivocally establish the role of the 'linker' region in ARE's molecular mechanism. Second, the low efficiency of protection by VgaA LC in the reconstituted system indicates that the biochemical experiments do not reveal the full story. One possible explanation could be that in the test tube, the dislodged antibiotic rapidly rebinds to the ribosome, thus compromising VgaA LC -mediated protection. However, this is unlikely, since in direct competition experiments the puromycin test reaction is sufficiently fast to kinetically outcompete the slowly binding lincomycin (Supplementary Figure S2B). A second explanation is that the reconstituted system is missing one or more components or co-factors crucial for the full activity of VgaA LC . It is tempting to speculate that in order to achieve efficient removal of the dislodged antibiotic, the activity of the ABCF ARE is coordinated with a direct interaction of the ARE with an efflux pump. Experiments with Streptococcus pneumoniae macrolide resistance ARE MsrD indicate a possible candidate. The msrD gene is co-transcribed with another ORF encoding MefE (macrolide efflux E) efflux pump, a member of major facilitator superfamily (56). While acting alone in the absence of MefE, the MsrD ABCF confers only low-level macrolide resistance (57), but the two proteins synergize when acting together (58). Although they are Streptococcal resistance factors, MefE and MsrD confer macrolide resistance in E. coli, suggesting that the proteins are functional when expressed heterologously (59). While GFP-labelled MsrD localized in the E. coli cytoplasm, GFP-labelled MefE is localized at cell poles--but only when co-expressed with wild type MsrD. This indicates a physical interaction between the two proteins. However, just how general this kind of interaction might be among ABCF AREs is unknown. Genetic, microbiological and molecular biology experiments with Staphylococcal VgaA are the next step for identification of its potential interactive partners, paving the way for more refined biochemical and structural studies.