Tobramycin variants with enhanced ribosome-targeting activity

Abstract

With the increased evolution of aminoglycoside (AG)-resistant bacterial strains, the need to develop AGs with (i) enhanced antimicrobial activity, (ii) the ability to evade resistance mechanisms, and (iii) the capability of targeting the ribosome with higher efficiency, is more and more pressing. The chemical derivatization of the naturally occurring tobramycin (TOB) by attachment of 37 different thioethers groups at the 6″-position led to the identification of generally poorer substrates of TOB-targeting AG-modifying enzymes (AMEs). Thirteen of these displayed better antibacterial activity than the parental TOB while retaining ribosome-targeting specificity. Analysis of these compounds in vitro shed light on the mechanism by which they act and revealed three with clearly enhanced ribosome-targeting activity.

Aminoglycosides (AGs) represent one of the major groups of antibiotics that target the bacterial ribosome. They interfere with translation,^[1] promote errors during decoding,^[2] and inhibit translocation^[3] and ribosome recycling.^[4] AGs have been shown to target the decoding site of the 30S subunit, interacting at the internal loop of helix h44 in which A1408 lies across from A1492-A1493.^[5] The structurally most-conserved portion of the AGs (rings I and II) forms most of the contacts to h44. Ring I intercalates into h44, stacking on G1491 and forming hydrogen bonds with A1408. This occludes A1492 and A1493 from within h44 and hence stabilizes a “flipped out” conformation of these nucleotides. An analogous rearrangement occurs upon codon recognition during decoding: A1492 and A1493 flip out of h44 and dock into the minor groove of the codon-anticodon helix.^[6] The ability of AGs to stabilize this conformation of h44 thus appears to disturb translational fidelity during protein synthesis. AG binding pays the energetic cost of the rearrangement and thereby stabilizes tRNA in the A site. This leads to miscoding, because near-cognate aa-tRNA is also stabilized.^[2] This also inhibits translocation, because A-tRNA cannot readily move from the A- to the P-site.^{[3, 7]}

The clinical usefulness of AGs has been seriously compromised by the growing prevalence of various resistance mechanisms among pathogenic bacteria. These mechanisms include the decrease in AG uptake into the bacteria, the alteration of the bacterial ribosome, and the acquisition of AMEs, which represent the major cause of resistance to AGs.^[8] With more than 100 AMEs identified, these enzymes pose a serious health threat as they chemically alter the structures of AGs by N-acetylation (AACs), O-phosphorylation (APHs), or O-nucleotidylation (ANTs). To overcome this issue, AGs that could evade the action of AMEs, while still targeting bacterial ribosomal RNA, have been investigated. This has led to the development of structurally constrained AGs that would mimic the ribosome-bound AG conformation,^[9] guanidinylated AGs,^[10] and AG dimers,^[11] which have been shown to also bind viral and human RNAs.^[12]

We have previously synthesized a number of 6″-thioether TOB variants and assayed their antimicrobial activities.^[13] Many of these compounds exhibited bacteriolytic activity, raising the possibility that the mode of action in these cases involves membrane disruption rather than translation inhibition. Herein, we present the synthesis of 18 additional TOB variants (Scheme 1), establish their antibacterial activity profile, and investigate the mechanism by which the 18 novel and the 19 previously reported 6″-thioether TOB variants inhibit bacterial growth, using AG-resistant ribosomes with mutations in the primary helix h44 site.

Scheme 1.

Synthesis of 6″-thioether TOB variants.

TOB was modified at the 6″-position with various thioether groups (Scheme 1). While compounds 3a–j, l, m, p–t, jj, and kk have been described previously,^[13–14] compounds 3k, n, o, and u–ii are new. Our collection encompasses a diverse set of 6″-substituents, including linear, branched, and cyclic alkyl groups, and substituted aromatic rings.

Compounds 3a–f, k–p, and u–jj were screened for their antibacterial activity against 19 diverse bacterial strains and their minimum inhibitory concentrations (MICs) were determined (Table S1). Among the variants with aliphatic substituents, compound 3f (with a C₁₄ chain) is the most potent against several of the TOB-resistant bacterial strains, including the newly tested E. faecium (C) and S. pyogenes (L). Compounds bearing aromatic substituents generally exhibited promising antibacterial activity (MIC 9.4 μg/mL) against B. anthracis (A), B. subtilis (B), L. monocytogenes (E), M. smegmatis (I), S. aureus NorA (J), S. epidermidis (K), E. coli (N), H. influenzae (H), and S. flexneri (S). Several of these, including compounds 3m–n, p, x–z, aa, cc–dd, ff, gg, and ii were even capable of inhibiting the growth of various bacterial strains at an MIC 2.4 μg/mL. Furthermore, additional substitution at the ortho- or para-positions on the aryl ring appeared to improve antibacterial activity. It is worth mentioning that compounds 3m (MIC ≤2.4 μg/mL), 3p (MIC 1.2 μg/mL), and 3jj (MIC ≤9.4 μg/mL) displayed a 16- to 128-fold drop in MIC value compare to TOB (MIC >150 μg/mL) against E. coli TolC (O), while compound 3z (MIC 18.8 μg/mL) was 8 times more active than TOB (MIC 150 μg/mL) against M. intracellulare (G).

Since these compounds demonstrated comparable or better potency against E. coli strains, we further investigated their mechanisms of action. We compared their abilities to inhibit growth of E. coli Δ7 prrn containing wild-type (WT), A1408G, or G1491U ribosomes. Strain Δ7 prrn lacks all seven chromosomal rRNA operons and instead contains a single plasmid-borne rRNA operon.^[15] Hence, each of these strains contains a homogeneous population of WT or mutant ribosomes. Mutations A1408G and G1491U target the primary AG binding site of helix h44 of the 30S subunit, and each mutation confers resistance to a number of AGs.^[16]

TOB inhibited E. coli Δ7 prrn WT (MIC 18.8 μg/mL) and failed to inhibit either Δ7 prrn A1408G or G1491U (MIC >150 μg/mL) (Table S2), indicating that TOB inhibits growth by binding its h44 site, consistent with previous results.^{[16a, 16c]} Many of the TOB variants showed a similar activity profile against these strains, inhibiting growth in an A1408- and G1491-dependent manner. Several of these (3m–n, p, x–z, aa–dd, ff–hh), all carrying an aryl ring substituent, are more potent than TOB and retain target specificity.

Two compounds (3f and 3g) were found to have indistinguishably strong antibacterial activity against Δ7 prrn WT, A1408G, and G1491U, showing that these compounds act through a distinctmechanism, independent of h44. These compounds, substituted with linear alkyl chains (C₁₄ and C₁₆, respectively) were previously found to have cellulolytic activity,^[13] which may be their primary mechanism of action. Compounds 3d and 3e, with slightly shorter alkyl chains (C₁₀ and C₁₂, respectively), have lower activity with little-to-no h44-dependence.

Finally, a number of compounds were found to be virtually inactive against all the tested strains. These include compounds 3b, c, h, and i; substituted respectively with C₆, C₈, C₁₈, or C₂₂ linear alkyl chains; compound 3s, modified with 4-t-butyl thiophenyl group; and compound 3kk, carrying a 4-methylcoumarin-7-yl group.

Some AGs are potent inhibitors of EF-G-dependent translocation.^{[3, 7, 17]} This is believed to be due to the stabilization of A-site tRNA that occurs when these antibiotics bind their primary site in h44. We suspected that TOB too would inhibit translocation, allowing effects of the 6″-substituents to be compared in vitro. To test this, we purified WT and A1408G ribosomes from the corresponding Δ7 prrn strains and used toeprinting to measure the extent of translocation in the presence of various concentrations of TOB, and the data obtained were fit to the Hill equation. TOB strongly inhibited translocation of WT ribosomes (IC₅₀ 16 μM) (Fig. 1, Table 1). Interestingly, the Hill coefficient derived from the curve-fitting was consistently less than 1 (0.5–0.7), raising the possibility that TOB binds its primary (h44) and secondary (H69)^[5a] sites in a negative cooperative manner. With A1408G ribosomes, TOB had substantially reduced potency (IC₅₀ 700 μM), and the data could be well fit with a Hill coefficient of either 1 or >1, depending on the particular experiment. These data show that inhibition of translocation by TOB normally depends on the primary h44 site, and mutation of that site qualitatively changes the concentration dependence of inhibition.

Figure 1.

Effects of TOB variants on ribosomal translocation. The extent of EF-G-dependent translocation was measured in control (WT; circles) and mutant (A1408G; squares) E. coli ribosomes with various concentrations of TOB variants.

Table 1.

Translocation inhibition (IC₅₀ values in μM) activities of TOB variants in WT or mutant A1408G E. coli ribosomes.

AG	Control (WT)	A1408G
TOB	16	700
3d	>1000	ND
3e	>1000	ND
3h	>1000	ND
3k	330	660
3l	230	>1000
3m	44	>1000
3n	5	250
3o	47	430
3p	36	>1000
3u	100	620
3v	100	990
3w	<1; 360a	650
3x	<1; 210a	310
3y	72	210
3z	<1; 280a	350
3aa	7.7	370
3bb	19	140
3cc	<5; 290a	240
3dd	1.2	230
3ee	33	180
3ff	<1; 250a	320
3gg	39	130
3hh	66	270
3ii	350	410
3jj	11	100

^aCurves in these cases exhibited complex concentration dependence. Data were fit to the sum of two Hill functions, and the reported values correspond to the two deduced inflection points (IC₅₀ values).

ND = not determined.

Next, the effects of several TOB variants on translocation were analyzed (Fig. 1, Table 1). Many behaved similarly to the parental compound, although differences correlating with the structure of the substituent were observed. Four compounds inhibited translocation more strongly than TOB in both WT and A1408G ribosomes. Three of these (3n, aa, dd) have ortho-substituted thioaryl groups, whereas the fourth carries a naphthyl moiety (3jj). For the former compounds, IC₅₀ values were similarly decreased for WT and mutant ribosomes. Compound 3jj, on the other hand, exhibited some loss of specificity, as the ratio of IC₅₀ values (WT vs A1408G) dropped from 44 to 9. Compounds 3u and 3x (with o-methoxythiophenyl or o-fluorothiophenyl groups) were notably poorer inhibitors of translocation.

Compounds with meta-substituted thioaryl groups (3o, v, y, bb, ee) were generally less potent translocation inhibitors than their ortho-substituted counterparts and had reduced specificity for WT ribosomes (Fig. 1, Table 1). As with ortho-substituted set, the methoxythiophenyl and fluorothiophenyl modified compounds (3v and y) were the weakest inhibitors of the group.

Many of the compounds with para-substituted aryl rings (3w, z, cc, ff) gave complex inhibition curves that failed to fit the standard Hill equation (Fig. 1). In these cases, the lowest concentration of drug tested (10 μM) resulted in a substantial degree of inhibition, and further inhibition occurred only gradually over the higher range of drug concentrations. While the basis of this complexity remains unclear, although highly speculative, it may arise from two distinct populations of ribosomal complexes, one being considerably more sensitive to AG inhibition than the other. Accordingly, the data were fit to the sum of two Hill functions, yielding inflection points (i.e., IC₅₀ values) for the two putative distinct populations (Table 1). In the context of the cell, potent inhibition of even a small subset of translating ribosomes would have deleterious consequences, consistent with the enhanced biological activity of these variants (Table S2). Compound 3x, which carries an ortho-fluorothiophenyl substituent, showed a similarly complex inhibition curve (Fig. 1). Finally, compounds 3d, e, and h, with linear alkyl chains, failed to inhibit translocation in vitro (Fig. 1, Table 1).

Compounds 3n, aa, and dd, with ortho-substituted aryl groups, thus appear to be more potent inhibitors of both bacterial growth and ribosomal translocation than TOB in E. coli strains. In these cases, inhibition of ribosomes in vivo and in vitro remains strictly dependent on A1408, showing that higher potency comes without loss of target specificity. These TOB variants thus bind the primary h44 site with increased affinity. How these 6″-aryl substituents promote TOB binding remains to be determined. Co-crystal structures of ribosomes bound to related AGs provide some clues to the basis of enhanced binding.^[5a] Rings I and II of these TOB variants make similar sets of contacts to nucleotides 1407–1409 and 1491–1495 of helix h44. Ring III of gentamicin contacts nucleotides 1405–1407, forming hydrogen bonds with G1405 and C1407. Presumably, the analogous ring III of TOB occupies a similar position. The 6″-aryl groups of the TOB variants may form a favourable stacking interaction with a nearby rRNA nucleotide, such as C1404, and thereby stabilize binding.

For the linear alkyl substituted compounds, potency and mechanism of action change as a function of chain length. Compound 3a, with the shortest chain (C₄) of the series, has biological activity indistinguishable from TOB, inhibiting cell growth in an A1408- and G1491-dependent manner. When the length of the chain is increased to C₆ (3b), antibacterial activity is completely lost (Table S2), presumably due to a steric clash with the ribosome, shedding light on the spatial constraints of the h44 site. Further lengthening of the alkyl chain to C₁₀ through C₁₆ (3d–g) restores antibacterial activity, although this activity is h44-independent. Indeed, these compounds do not appear to target the ribosome, as 3f fails to inhibit in vitro translation^[13] and 3e fails to inhibit translocation (Fig. 1). Rather, these compounds target the cell membrane and probably act via bacteriolysis.^[13] Compounds with even longer chains (3h, C₁₈; 3i, C₂₂) lose potency, likely due to diminished bacteriolytic activity.

As quite a number of our TOB variants demonstrated good antibacterial activity, we evaluated their susceptibility to AMEs, which greatly contribute to bacterial resistance to AGs. We measured the relative activities of four AMEs with our TOB variants and compared them to that of TOB itself (Fig. 2). While an increase in catalytic activity of AAC(3)-IV^[1a] and ANT(4′)-Ia^[18] was noticeable in the majority of TOB variants with aliphatic substituents (3a–c, f, k, l), in general, TOB variants bearing an aromatic ring appeared to be poorer substrates of these AMEs. This implies that TOB variants with aliphatic substituents were more susceptible to modifications by AAC(3)-IV and ANT(4′)-Ia than TOB variants bearing an aromatic ring. In light of the information gathered from the crystal structures of various AMEs with our parent drug TOB (e.g., AAC(2′)-Ic (PDB: IM4D),^[19] APH(2″)-IVa (PDB 3SG8),^[20] Eis (PDB: 4JD6),^[21] along with that of AMEs with other AGs (e.g., AAC(2′)-Ic (PDB: IM4I),^[19] AAC(6′)-Ib (PDB: 1V0C),^[22] APH(2″)-IVa (PDB: 3SG9),^[20] APH(2″)-Id/APH(2″)-IVa (PDB: 4DFB),^[23] and APH(3′)-IIIa (PDB: 1L8T)^[24] with kanamycin A; AAC(2′)-Ic (PDB: IM4G),^[19] AAC(6′)-Iy (PDB: 1S3Z),^[25] and AAC(6′)-Ib (PDB: 2BUE)^[22] with ribostamycin; AAC(6′)-Ib (PDB: 2VQY),^[22] and Eis (PDB: 4QB9)^[26] with paromomycin; APH(3′)-IIIa (PDB: 2B0Q)^[24] with neomycin B), it appears that the AG binding pocket of AACs is very hydrophilic, lined with several water molecules and amino acid residues with acidic side chains. Replacement of the hydroxyl group at the 6″-position with a hydrophobic moiety might thus weaken the interactions of the resulting AGs in the binding pocket, reducing their susceptibility to AMEs. Taking all of our combined data into consideration (Figs. 2 and 3, Tables 1 and S1), compounds 3n, aa, and dd, which all contain an ortho-substituted thioaryl group, exhibited the most potent inhibitory activity against bacterial growth and ribosomal translation, and were also among the poorest substrates for all the tested AMEs. These variants could therefore evade the action of TOB-targeting AMEs AAC(6′)-Ie/APH(2″)-Ia,^[1a] AAC(6′)-Ib′,^[27] and ANT(4′)-Ia better than TOB. It is important to note that the most common AAC(6′) enzymes were the least active against all variants tested, a highly encouraging result.

Figure 2.

Bar graph showing the relative initial rates of reactions of the listed AMEs with variants 3a–jj. Rates are normalized to TOB. * indicates that 3f with AAC(6′)-Ib′ had activity >200% and is not shown here.

In conclusion, 18 novel 6″-thioether TOB variants were synthesized and, together with the 19 previously reported, their antibacterial activity were evaluated. Compound 3f (with a C₁₄ chain) exhibited the most potent antibacterial activity; meanwhile compounds bearing aromatic substituents were more active against E. coli strains. Importantly, we identified several TOB variants with enhanced ribosome-targeting activity. Compounds 3n, aa, and dd inhibited bacterial growth and ribosomal translation better than TOB in E. coli. These compounds also served as poor substrates for many AMEs, including those known to target parental TOB, suggesting that AMEs will have little-to-no effect on these AG variants. As the addition of aromatic moieties to the 6″-position of TOB appears to be a promising avenue for enhancing ribosome-targeting activity, studies aimed at adding aromatic substituents to AGs at various positions are currently underway in our laboratories.