A laboratory-based predictive pathway for the development of Neisseria gonorrhoeae high-level resistance to corallopyronin A, an inhibitor of bacterial RNA polymerase

ABSTRACT The continued emergence of Neisseria gonorrhoeae strains that express resistance to multiple antibiotics, including the last drug for empiric monotherapy (ceftriaxone), necessitates the development of new treatment options to cure gonorrheal infections. Toward this goal, we recently reported that corallopyronin A (CorA), which targets the switch region of the β′ subunit (RpoC) of bacterial DNA-dependent RNA polymerase (RNAP), has potent anti-gonococcal activity against a panel of multidrug-resistant clinical strains. Moreover, in that study, CorA could eliminate gonococcal infection of primary human epithelial cells and gonococci in a biofilm state. To determine if N. gonorrhoeae could develop high-level resistance to CorA in a single step, we sought to isolate spontaneous mutants expressing any CorA resistance phenotypes. However, no single-step mutants with high-level CorA resistance were isolated. High-level CorA resistance could only be achieved in this study through a multi-step pathway involving over-expression of the MtrCDE drug efflux pump and single amino acid changes in the β and β′ subunits (RpoB and RpoC, respectively) of RNAP. Molecular modeling of RpoB and RpoC interacting with CorA was used to deduce how the amino acid changes in RpoB and RpoC could influence gonococcal resistance to CorA. Bioinformatic analyses of whole genome sequences of clinical gonococcal isolates indicated that the CorA resistance determining mutations in RpoB/C, identified herein, are very rare (≤ 0.0029%), suggesting that the proposed pathway for resistance is predictive of how this phenotype could potentially evolve if CorA is used therapeutically to treat gonorrhea in the future. IMPORTANCE The continued emergence of multi-antibiotic-resistant strains of Neisseria gonorrhoeae necessitates the development of new antibiotics that are effective against this human pathogen. We previously described that the RNA polymerase-targeting antibiotic corallopyronin A (CorA) has potent activity against a large collection of clinical strains that express different antibiotic resistance phenotypes including when such gonococci are in a biofilm state. Herein, we tested whether a CorA-sensitive gonococcal strain could develop spontaneous resistance. Our finding that CorA resistance could only be achieved by a multi-step process involving over-expression of the MtrCDE efflux pump and single amino acid changes in RpoB and RpoC suggests that such resistance may be difficult for gonococci to evolve if this antibiotic is used in the future to treat gonorrheal infections that are refractory to cure by other antibiotics.

Since the mid-1930s, antibiotic therapy has been the mainstay for curing gonococcal infections and for reducing the spread of this STI in the community (2).Despite the overall effectiveness of antibiotic treatment regimens over the past decades, Ng strains expressing resistance to each new antibiotic brought into clinical practice have ultimately emerged, prompting changes in therapeutic regimens.Currently, ceftriaxone (Cro) is the last remaining antibiotic for empiric monotherapy; however, Cro-resistant strains have emerged, and some of these are also spreading internationally (3).
In the absence of a protective vaccine, the development of new, effective, and affordable antibiotics for the treatment of gonorrhea is essential.Toward that end, we recently reported that the naturally produced antibiotic, corallopyronin A (CorA), exhibits anti-gonococcal activity (4).CorA binds to the pocket, in the vicinity of the switch-2 region of the β′ subunit (RpoC), of the bacterial DNA-dependent RNA polymerase (RNAP) (5,6).CorA-induced refolding of the switch-2 structure, which in its natural (antibioticfree) conformation is essential for transcription initiation, inhibits RNAP activity (5) and has antibacterial action against several pathogens (7,8).Recent technical advances in CorA production and purification have stimulated renewed interest in developing it as a clinically useful antibiotic (7,9).The pre-clinical development of this natural compound has intensified (10), and oral formulations for use in humans have been developed (11).
Given the global need for new antibiotics to treat gonorrhea, we initiated studies to test CorA effectiveness against Ng.In our initial studies, we found that CorA was active against a panel of gonococcal strains [minimal inhibitory concentrations (MICs) ranging from 0.125 to 2.0 µg/mL], including those expressing resistance to different antibiotics, such as rifampin, which also binds to RNAP but at a region distinct from CorA, as well as Cro and azithromycin (4).Moreover, CorA was effective in killing gonococci growing in a biofilm state and when interacting with primary human cervical epithelial cells.Important for the continuation of pre-clinical studies, we examined the possibility that gonococci could develop CorA resistance, and results showed that spontaneous mutants expressing at least a fourfold increase in the CorA MIC could not be recovered, thus indicating a very low frequency of spontaneous CorA resistance (frequency of ≤10 −10 ).However, at a lower level of selection (2× MIC), a spontaneous mutant was recovered that displayed a fourfold increase in the CorA MIC when compared to the parent strain, FA19.This mutant had a single amino acid change at position 27 of the MtrR regula tory protein, which normally represses expression of the mtrCDE-encoded efflux pump operon, and this mutation was found to be responsible for the elevated CorA MIC.This mtrR-27 mutation resulted in increased expression of the mtrCDE-encoded efflux pump, suggesting that it reduced MtrR repressive action on mtrCDE transcription.Importantly, loss of MtrCDE efflux pump activity was found to reduce the CorA MIC, indicating that this antibiotic is a substrate for this antimicrobial export system (4).Thus, MtrCDE efflux of CorA was invoked as a mechanism by which gonococci could reduce their susceptibil ity to CorA.We, therefore, asked if other mutations providing higher levels of resistance (MIC of ≥32 µg/mL), such as amino acid changes in the RNAP, as is observed with other bacteria (12,13), could develop spontaneously in Ng.We now report that high-level CorA resistance can develop in gonococci, but that this requires a multi-step process involving an over-expressed MtrCDE efflux pump in combination with single amino acid changes in two subunits (RpoB and RpoC) of RNAP.

RESULTS AND DISCUSSION
To determine whether gonococci could achieve high levels of CorA resistance in a single step, as has been reported for Staphylococcus aureus (12,13), we selected for spontaneous mutants of strain FA19 mtrR-27 (CorA MIC = 0.5 µg/mL; Table 1) at 2× the parent MIC.We recovered mutants with this selection at a frequency of 5 × 10 −8 .The CorA MIC against three randomly chosen mutants was determined by agar dilution and found to be 8 µg/mL, representing a 16-fold increase in resistance when compared to the parental (mtrR-27) strain.Whole genome sequencing revealed that these mutants contained a single missense mutation at codon 347 of rpoC (rpoC-347), which encodes the β′ subunit of RNAP, resulting in a lysine to arginine change at amino acid 347 of RpoC.Transformation of strain FA19 mtrR-27 with a PCR (polymerase chain reaction) product encompassing the entire rpoC-347 gene showed that transformants phenocopied the donor CorA MIC and that the (K347R) mutation was present.We modeled this amino change for the interaction of RpoC with CorA using a previously described structure of CorA interacting with RpoB and RpoC from Wolbachia endobacteria (13) and surmised that the RpoC Lys-347 makes direct contact with CorA (see below).
We next used strain FA19 mtrR-27 rpoC-347 to select for spontaneous mutants at 2× the CorA MIC.Colonies were recovered (frequency of 3.8 × 10 −8 ); the CorA MIC of three selected mutants was 32 µg/mL.Whole genome sequencing revealed that one selected mutant had a single missense mutation at codon 1328 of rpoB that would result in a leucine to serine change.Interestingly, modeling studies (see below) suggested that RpoB L1328 does not make direct contact with CorA, suggesting that any influence this mutation might have on the interaction with CorA is allosteric (Fig. 1).Based on this inference, we tested if this rpoB mutation could confer the observed CorA resist ance phenotype by transforming the parent strain with an intact mutant rpoB gene.Three putative transformants were selected for further study, and their rpoB genes were PCR-amplified.DNA sequencing of these PCR products showed that all contained the donor mutation at codon 1328 but also had other single missense mutations at  codons 1325 (valine to isoleucine), 1329 (glutamic acid to lysine), or 1376 (leucine to tryptophan).Unlike the serine at position 1328, all three of the wild-type amino acids at these positions are predicted to directly contact CorA, which is described in detail below (Fig. 2A).
To determine if CorA resistance in gonococci could develop due to mutations in rpoB independent of the leucine to serine change at amino acid position 1328, we focused on the neighboring glutamic acid residue at position 1329.For this, we constructed transformants of strains FA19 mtrR-27 and FA19 mtrR-27 rpoC-347 that contained a site-directed mutation at position 1329.The presence of the mutation was confirmed in selected transformants.Importantly, the constructed strains differed significantly in their levels of CorA resistance with the co-residence of the rpoC-347 mutation conferring the highest CorA MIC (256 µg/mL; Table 1).
The dual presence of the rpoB-1329 and rpoC-347 alleles could confer high-level CorA resistance; therefore, we questioned whether they are present in global clinical strains of Ng.We initially screened the whole genome sequences of 1,089 gonococcal strains (see Materials and Methods) for the number and type of rpoB and rpoC alleles.This screen revealed the presence of 46 RpoB and 49 RpoC amino acid sequence allelesnone contained the RpoB and RpoC mutations described above.We then expanded this analysis by interrogating a global collection of 33,256 Ng genomes.This analysis did not detect the RpoB variants at amino acid positions 1328, 1329, or 1376 or an RpoC variant at amino acid position 347; however, one strain encoded an RpoB variant at position 1325 (valine to isoleucine).Collectively, this analysis showed that RpoB and RpoC CorA resistance determinants are rare (≤0.0029%) in contemporary global Ng strains.
Taken together with our previous study (4), the results presented herein suggest that high-level CorA resistance in gonococci could emerge through a multi-step process involving over-expression of the MtrCDE efflux pump in combination with specific, step-wise, amino acid replacements in RpoB and RpoC.To ascertain whether the MtrCDE efflux pump contributes to high-level CorA resistance, we constructed a genetic derivative of strain FA19 mtrR-27 rpoC-347 rpoB-1329 containing an insertionally inactivated mtrD gene (FA19 mtrR-27 rpoC-347 rpoB-1329 mtrD::kan), which encodes the inner membrane transporter protein (MtrD) of the MtrCDE efflux pump (2,15).A representative transformant was fourfold more susceptible to CorA than the FA19 mtrR-27 rpoC-347 rpoB-1329 parent strain, indicating that an active MtrCDE efflux pump is needed for high-level CorA resistance (256 µg/mL) in gonococci (Table 1).

Modeling-based implications of RpoB and RpoC amino acid changes on Ng CorA resistance
In the RNAP/CorA structural model (6,14), the RpoC Lys347 side chain adopts an extended conformation, which is stabilized by the hydrogen bond with the main chain carboxyl of RpoB Leu1341, and stacks on the central ring of CorA (Fig. 2A).The bulkier side chain of Arg in the RpoC K347R variant would not be able to adopt this extended conformation because of the steric clash with the adjacent RpoB structures (Fig. 2B).The Arg side chain, therefore, should be bent, most likely toward the open space in the CorA binding pocket, since bending in another direction (though also possible upon certain structural rearrangements) is largely restricted by the surrounding protein side chains.The bent Arg side chain in this position would, therefore, sterically prevent CorA binding (Fig. 2C).
RpoB Leu 1328 forms no interactions with CorA and is located distantly to the CorA binding site (minimal distance ~10 Å) (Fig. 1A).The RpoB Leu 1328 side chain is part of a hydrophobic core within a structural domain formed by the RpoB and RpoC residues in the vicinity of the CorA binding site.Some residues from this domain (in particular, RpoB Val 1325 and RpoB Glu 1329) form a part of the CorA binding site and directly interact with this antibiotic (Fig. 2A).In the absence of direct contact with CorA, it seems that the RpoB L1328S variant may exert only an allosteric effect on CorA binding.The smaller Ser side chain (versus that of Leu) would be unlikely to sterically disturb the hydrophobic core that could have resulted in the significant conformational changes and reconfiguration of the CorA binding site.In fact, this substitution opens a free space within the hydrophobic core (Fig. 1B).This open space may trigger a shift of other hydrophobic residues in this core toward the center to occupy this space and re-establish an optimal network of hydrophobic interactions.Our modeling suggests that RpoB Phe 1320 is a plausible candidate for such a shift (Fig. 1).The main chain of Phe 1320 forms a hydrogen bond with the main chain of RpoC Arg 347 in the double mutant.It is, therefore, possible that the RpoB 1328-induced alterations of the hydrophobic core transmit an allosteric signal through the main chain interactions to RpoC Arg 347 to further restrict its bent conformation, in which its side chain sterically clashes with CorA.Thus, it seems that the role of the RpoB L1328S mutation might be to enhance the steric effect of RpoC Arg 347 on CorA binding.
The RpoB mutated residues, which complement the RpoB L1328S mutation in the double mutant variants (V1325I, E1329K, and L1376W), directly interact with CorA in its binding pocket (Fig. 2A).Therefore, their effect on antibiotic resistance is likely achieved through steric hindrance with CorA.The modeling of these mutants could be performed with high confidence since the alternative, possible sterically favorable, conformations of their side chains are restricted by the surrounding protein residues.From a structural perspective, it seems that each of them alone (without the RpoB L1328S substitution) may account for the significant MIC effects.This is consistent with the results for the RpoB E1329K mutant (Fig. 2D and E; Table 1).In the double mutants, their role might be in enhancing the potentially modest allosteric effect of the RpoB L1328S substitution.
Historically, with every new antibiotic brought into clinical practice to treat gonor rheal infections since the mid-1930s with sulfonamides, resistance in Ng has developed rapidly (sulfonamides) or slowly (penicillin) by spontaneous mutation or horizontal gene exchange or both (2), resulting in their removal from recommended treatment regimens.Emphasizing this point, it is noteworthy that just in the past few decades, Ng resistance to newly introduced antibiotics has occurred with fluoroquinolones, azithromycin, and extended-spectrum cephalosporins.With Cro as the last remaining antibiotic for empiric monotherapy now employed in many countries and Cro resistance spreading interna tionally (3), it is essential that new, effective, and affordable antibiotics be introduced into the clinic globally.
We previously showed that CorA was effective in killing multidrug-resistant Ng strains (MIC range of 0.125-2 µg/mL) and could exert activity against Ng growing as a biofilm on human epithelial cells or an abiotic surface (4).Importantly, this RNAP-bind ing antibiotic was active against Ng (strain WHO M) exhibiting high-level resistance to rifampin, another RpoB-targeting antibiotic, and cross-resistance was not observed.Notably, unlike earlier studies with S. aureus (12), we could not identify single-step, spontaneous CorA-resistant mutants (as defined by an MIC of ≥32 µg/mL).However, using antibiotic-susceptible strain FA19 (CorA MIC = 0.125 µg/mL) in spontaneous mutant selection studies, we found that a mutation in the mtrR gene (mutant allele termed mtrR-27) that results in over-expression of the MtrCDE efflux pump decreased Ng susceptibility to CorA by fourfold (MIC = 0.5 µg/mL).Wild-type MtrR represses expression combination with the allosteric RpoB L1238S mutant (lacking direct contacts with CorA) that confer significant resistance to CorA, are highlighted within black boxes with respective substitutions outlined in magenta.(B, C) Structural modeling of the RpoC (β′) K347R mutant.In the optimal extended conformation observed in the X-ray structure (14) for the Lys side chain, the bulkier Arg side chain forms unfavorable close contacts with the adjacent residues of RNAP (B).In the bent, sterically allowed conformation that avoids close contacts of Arg with the RNAP residues, its side chain clashes with the CorA backbone in the modeled RNAP/CorA complex (C).(D, E) Structural modeling of the RpoB (β) E1329K mutant.In the wild-type enzyme, the Glu side chain forms two hydrogen bonds with the N 1 and O 5 atoms of CorA (refer to panel A), likely stabilizing the complex structure (D).In the E1329K mutant, the bulkier side chain of Lys would likely produce steric hindrance with the CorA backbone, which might be enhanced by the repulsive interactions of the Lys NH 2 group with the N 1 atom of CorA (E).In the structural panels (B-E), the protein structure is represented by the Cα trace with the side chains.The RpoB (β) and RpoC (β′) subunits of RNAP, and the CorA structure are shown as the yellow, green, and white balls-and-sticks models, respectively.The mutated residues are emphasized by cyan color.The close contacts (<2.3 Å) and hydrogen bonds are shown in solid magenta and dashed red lines, respectively. of the mtrCDE operon, but multiple cis-or trans-acting mutations that negate mtrR expression or DNA-binding ability have been identified and are present in nearly 50% of contemporary Ng global strains [reviewed in reference (15)].Furthermore, because loss of the MtrD transporter component of the pump increased Ng susceptibility to CorA, we used our mtrR-27 strain in this study as a host for the selection of additional mutants expressing high levels of CorA resistance.Based on this strategy, we propose a model by which a combination of rpoB and rpoC mutations would be required for Ng to express a CorA MIC of ≥32 µg/mL.This requirement for mutations impacting both RpoB and RpoC structure suggests that it may be difficult for Ng, unlike S. aureus, to achieve CorA resistance in a single step.Notably, the RpoB-1329 and RpoC-347 mutant proteins would have direct effects on CorA binding (Fig. 2), whereas the RpoB-1328 mutant protein (observed in the selection system used herein) would likely have an allosteric impact on other CorA interacting amino acids within its binding site in RNAP (Fig. 1).Strikingly, except for a single strain (isolated in the United Kingdom) with an rpoB-1325 allele, all 33,256 Ng genomes examined were predicted to have wild-type amino acids at positions 1328, 1329, and 1376 of RpoB.Based on the rare occurrence of the rpoB-1325 mutation in global strains we asked what the consequence of this mutation would be with respect to CorA resistance.In this regard, a transformant of strain FA19 mtrR-27 rpoC-347 containing a site-directed mutation at codon 1325 was only twofold more CorA-resistant (MIC = 16 µg/mL) than the recipient (MIC = 8 µg/mL).We conclude that, at present, the mutations in rpoB and rpoC identified herein that result in a CorA MIC of 32 µg/mL are extremely rare in global clinical Ng strains (estimated at ≤0.0029%).Taken together, we propose a multi-step pathway for how Ng CorA resistance might develop.This laboratory predictive pathway involves single amino acid changes in both RpoB and RpoC, and that the highest level of resistance (MIC of 256 µg/mL) identified in this study would require an active and over-expressed MtrCDE efflux pump (Table 1).

Ng strains and growth conditions
The Ng strains used in this study are listed in Table 1.Gonococci were grown on gonococcal base (GCB) agar (Difco, Sparks, MD, USA) containing Kellogg's supplements I and II at 37°C under a 5.0% (vol/vol) CO 2 -enriched, humid atmosphere (4).Liquid cultures of gonococci for growth assays were begun by inoculating plate-grown cells in pre-warmed GCB broth containing Kellogg's supplements I and II and 0.043% (wt/vol) and sodium bicarbonate and grown in a 37°C water bath with shaking.

Production and purification of CorA
High quality (>91% pure), research grade CorA was produced at the Department of Microbial Drugs, Helmholtz Centre for Infection Research, Braunschweig, Germany using the heterologous producer strain, Myxococcus xanthus DK1622::pDPO-mxn116-Pvan-Tpase, as previously described (4,7,11).Stock CorA solutions were prepared by dissolving in 50% dimethyl sulfoxide and stored at −80°C until used.

MIC determination
CorA MICs were determined by the agar dilution method essentially as described previously (4).The inoculum consisted of 5 µL [approximately 5 × 10 5 colony forming units (CFUs)], which was obtained from overnight GCB agar cultures, as described above, resuspended in GCB broth to give approximately 10 8 CFU/mL.Plates were incubated for 24 hours [37°C, 5.0% (vol/vol) CO 2 ] before being photographed for data storage purposes.We considered fourfold or greater differences in CorA MIC values between isogenic strains as biologically significant, and reported values are representative from at least three independent assays.

Isolation of spontaneous mutants with increased MICs of CorA
Ng strain FA19 mtrR-27 was grown overnight on GCB agar, and heavy growth from five plates was collected and resuspended in 5 mL of GCB broth.Dilution plating onto GCB agar was performed to determine total colony forming units per milliliter, and 0.5 mL samples were plated onto GCB agar containing 2× or 4× the CorA MIC.Plates were incubated as described above for 48 hours before determining growth.

PCR amplification and sequencing of target genes
PCR amplification was performed as described previously (16,17) using primers listed in Table 2.Each of the three primer pairs was used to amplify rpoB and rpoC as follows: (i) rpoB1 and rpoB2, (ii) rpoB3 and rpoB6, and (iii) rpoB4 and rpoB5 for rpoB; and (i) rpoCF and rpoC4, (ii) rpoC1 and rpoC2, and (iii) rpoC5 and rpoCR for rpoC.All rpoB and rpoC PCR amplicons were sequenced using the same primer pair used for amplification.For amplification of mtrD::kan from strain KH14 (FA19 mtrD::kan), previously described oligonucleotide primers mtrD1 and mtrD2 were used (17; Table 2).All PCR products were purified with QiaQuick columns, also as described (16).Sequencing of PCR products was performed by Azenta Life Sciences (Burlington, MA, USA).

Site-directed mutagenesis and transformation
Site-directed mutagenesis of wild-type rpoB from strain FA19 was performed as described by Holley et al. (16) with modifications.Briefly, the 3′-end of rpoB was amplified by PCR using rpoB5 and rpoB7 primers (Table 2).The purified PCR product was then cloned into pBAD and used to transform Escherichia coli TOP10 as described (16).The rpoB insert from a resulting recombinant plasmid was sequenced using primers rpoB5 and rpoB7 to verify the presence of a wild-type sequence.To construct a single sitedirected mutation at amino acid position 1329, overlapping mutagenic oligonucleotide primers, rpoB1329F and rpoB1329R (Table 2), were used to amplify the entire plasmid construct using QuikChange Lightning Mutagenesis kit (Agilent Technologies, Santa Clara, CA, USA).Next, the PCR reaction was digested with Dpn1 and then used to transform E. coli XL-10 Gold ultra-competent cells.Transformants were selected on LB agar containing 100 µg/mL of ampicillin (Sigma Chemical Co., St. Louis, MO, USA).Plasmid DNA from a representative transformant was purified, and the insert was sequenced to verify the desired mutation.Using a plasmid containing the desired rpoB mutation, strains FA19 and FA19 mtrR-27 were transformed as described previously (16).Transformants were selected using CorA at 1-4× the MIC of the host strain.The presence of the rpoB1329 allele was confirmed by sequencing PCR products obtained from two putative transformants from each host strain.The complete rpoB and rpoC genes from transformants were sequenced, as described above, to confirm that, except for the rpoB1329 mutation, the rest of the sequences of both genes were wild type.

FIG 1
FIG 1 Structural model of the allosteric RpoB (β) L1328S mutant in the RNAP/CorA complex.(A, B) In the wild-type enzyme, RpoB L1328 (A) or its Ser substitution in the mutant (B) forms a part of the hydrophobic core of the domain adjacent to the CorA binding site.The residue in this position is located far away from the CorA binding site (the nearest distance is ~10 Å), suggesting that the L1328S variant likely possesses an allosteric effect on the CorA binding.The Leu to Ser amino acid substitution at position 1328 creates additional space in the hydrophobic core, where this substitution resides, potentially contributing to this effect.The protein structure is represented by the Cα trace with the side chains.The RpoB (β) and RpoC (β′) subunits of RNAP, and the CorA structure are depicted as the yellow, green, and white balls-and-sticks models, respectively.The wild-type Leu1328 residue (A) and its Ser substitution (B) are highlighted in cyan.The distances between Leu1328 or the Ser1328 mutant and CorA in the hydrophobic core are illustrated with cyan dashed lines.

FIG 2
FIG 2 Structural modeling of the steric mutants in the RNAP/CorA complex.(A) The schematic drawing illustrates the direct contacts between CorA and the RNAP residues in the model.RNAP amino acids from the RpoB (β) and RpoC (β′) subunits are depicted in light brown and green, respectively.Polar interactions (hydrogen bonds) and hydrophobic interactions are indicated by red arrows and dashed gray lines, respectively.RNAP mutations, which either alone or in (Continued on next page)

TABLE 1
Ng strains used in this study and their susceptibility to CorA a All MIC values are representatives of 3-5 determinations.b P.F. is for Philip Frederick and represents the first and middle names of Dr. Sparling.He provided the strain.

TABLE 2
Oligonucleotide primers used in this study